Wright State University Wright State University
CORE Scholar CORE Scholar
Browse all Theses and Dissertations Theses and Dissertations
2018
Impact of Noise Level on Task Performance and Workload and Impact of Noise Level on Task Performance and Workload and
Correlation to Personality Correlation to Personality
Kaylee Marie Eakins Wright State University
Follow this and additional works at: https://corescholar.libraries.wright.edu/etd_all
Part of the Operations Research, Systems Engineering and Industrial Engineering Commons
Repository Citation Repository Citation Eakins, Kaylee Marie, "Impact of Noise Level on Task Performance and Workload and Correlation to Personality" (2018). Browse all Theses and Dissertations. 1948. https://corescholar.libraries.wright.edu/etd_all/1948
This Thesis is brought to you for free and open access by the Theses and Dissertations at CORE Scholar. It has been accepted for inclusion in Browse all Theses and Dissertations by an authorized administrator of CORE Scholar. For more information, please contact [email protected].
Impact of Noise Level on Task Performance and Workload and Correlation to Personality
A thesis submitted in partial fulfillment
of the requirements for the degree of
Master of Science in Industrial and Human Factors Engineering
By
KAYLEE MARIE EAKINS
B.S.B.E., Wright State University 2017
2018
Wright State University
WRIGHT STATE UNIVERSITY
GRADUATE SCHOOL
April 19, 2018
I HEREBY RECOMMEND THAT THE THESIS PREPARED UNDER MY SUPERVISION
BY Kaylee Marie Eakins ENTITLED Impact of Noise Level on Task Performance and
Workload and Correlation to Personality TO BE ACCEPTED IN PARTIAL FULFILLMENT
OF THE REQUIREMENTS FOR THE DEGREE OF Master of Science in Industrial and
Human Factors Engineering
___________________________________
Mary Fendley, Ph.D., Thesis Director
___________________________________
Jaime Ramirez-Vick, Ph.D., Chair
Department of Biomedical, Industrial
and Human Factors Engineering
Committee on
Final Examination
___________________________________
Mary Fendley, Ph.D.
___________________________________
Frank Ciarallo, Ph.D.
___________________________________
Matthew Sherwood, Ph.D.
__________________________________
Barry Milligan, Ph.D.,
Interim Dean of the Graduate School
iii
ABSTRACT
Eakins, Kaylee Marie. M.S.I.H.E., Department of Biomedical, Industrial and Human Factors
Engineering, Wright State University, 2018. Impact of Noise Level on Task Performance and
Workload and Correlation to Personality. An ideal work environment supports a culture of high performance, low mental workload, and
quick turnarounds. The impact of noise on three types of tasks in a lab work environment were
examined while attempting to identify correlations between a subject’s personality and their
tolerance to noise. Neuroticism, agreeableness, conscientiousness, and extroversion correlated
significantly with subjective (NASA-TLX) and physiological mental workload measures (heart
rate variability and eye-tracking). The results show that task type impacts the performance, task
duration, and mental workload. Although the physiological workload measures showed
significant impact, the parameters standard deviation of R-R intervals and LF/HF ratio agreed
with the NASA-TLX scores while the parameters RMSSD value and standardized mean of R-R
intervals disagreed. Noise level nearly showed statistical significance with task duration and
LF/HF ratio; however, more research is necessary to completely rule out the influence of noise
level on the human participants.
iv
Table of Contents
1.0 Introduction ............................................................................................................................... 1
1.1 Background ........................................................................................................................... 1
1.2 Research Objective ................................................................................................................ 2
2.0 Literature Review...................................................................................................................... 2
2.1 Task and Multiple Resource Theory ..................................................................................... 2
2.2 Noise Interruption ................................................................................................................. 5
2.3 Mental Workload Analysis.................................................................................................... 6
2.3.1 NASA TLX..................................................................................................................... 6
2.3.2 Eye-tracking .................................................................................................................... 7
2.3.3 Heart Rate Variability ..................................................................................................... 9
2.4 Personality ........................................................................................................................... 12
3.0 Methods................................................................................................................................... 16
3.1 Experimental Design ........................................................................................................... 16
3.2 Participants .......................................................................................................................... 17
3.3 Stimuli and Apparatus ......................................................................................................... 18
3.4 Procedure ............................................................................................................................. 20
3.5 Data Analysis ...................................................................................................................... 21
3.5.1 Performance Scoring .................................................................................................... 21
3.5.2 Duration of task ............................................................................................................ 22
3.5.3 NASA-TLX .................................................................................................................. 22
3.5.4 Eye-tracking .................................................................................................................. 23
3.5.5 Heart Rate Variability ................................................................................................... 23
3.6 Hypotheses .......................................................................................................................... 24
3.6.1 Performance .................................................................................................................. 24
v
3.6.2 Duration of the Tasks ................................................................................................... 24
3.6.3 NASA-TLX Mental Workload ..................................................................................... 25
3.6.4 Eye-tracking and HRV Parameters ............................................................................... 25
4.0 Results ..................................................................................................................................... 26
4.1 Performance ........................................................................................................................ 27
4.2 Duration ............................................................................................................................... 30
4.3 NASA-TLX Mental Workload............................................................................................ 32
4.4 Physiological Mental Workload .......................................................................................... 34
4.4.1 Heart Rate Variability Analysis .................................................................................... 34
4.4.2 Eye-tracking Analysis................................................................................................... 39
4.4.3 Summary of Physiological Parameters ......................................................................... 43
4.5 Correlation Testing .............................................................................................................. 43
5.0 Discussion ............................................................................................................................... 48
5.1 Performance, Duration, and NASA-TLX ........................................................................... 48
5.2 Physiological Mental Workload .......................................................................................... 50
5.3 Correlation Tests ................................................................................................................. 55
6.0 Conclusion .............................................................................................................................. 56
7.0 Appendix ................................................................................................................................. 57
7.1 Appendix A: Experimental Design and Combinations Table ............................................. 57
7.2 Appendix B: Questionnaires and Task Problems (with answers) ....................................... 60
7.2.1 Noise Tolerance Questionnaire .................................................................................... 60
7.2.2 Big Five Inventory (Link) ............................................................................................. 61
7.2.3 NASA-TLX (Link) ....................................................................................................... 61
7.2.4 Anomaly Detection Task (with answers) ..................................................................... 62
7.2.5 Data Entry Task (with answers) ................................................................................... 62
vi
7.2.6 Mathematical Arithmetic Task (with answers) ............................................................ 63
7.3 Appendix C: Eye-tracking Illustrations............................................................................... 63
7.4 Appendix D: Residual Plots ................................................................................................ 67
7.4.1 Normal Distribution Checks ......................................................................................... 67
7.4.2 Residual vs. Predicted Plots.......................................................................................... 71
7.5 Appendix E: Connecting Letters Reports and Interaction Plots ......................................... 74
7.5.1 Task Performance ......................................................................................................... 74
7.5.2 Task Duration ............................................................................................................... 75
7.5.3 Mental Workload .......................................................................................................... 76
7.5.4 Heart Rate Parameters .................................................................................................. 77
7.5.5 Eye-tracking Parameters ............................................................................................... 79
7.6 Appendix F: Correlation Tables .............................................................................................. 80
8.0 References ............................................................................................................................... 93
vii
List of Figures
Figure 1: Illustration of R-R interval in EKG signal ...................................................................... 9
Figure 2: Graphical user interface for data entry task................................................................... 19
Figure 3: Anomaly detection task with anomaly circled .............................................................. 19
Figure 4: Arithmetic task set-up with manual pill counter, beads, and pill bottle ........................ 20
Figure 5: Average task type and noise level vs. performance (error bars are standard deviation) 28
Figure 6: Task type, noise level, and task type*noise level interaction connecting letters reports
for performance ............................................................................................................................. 29
Figure 7: Average of task duration vs. task type and noise level ................................................. 30
Figure 8: Task type, noise level, and task type*noise level interaction connecting letters reports
for task duration ............................................................................................................................ 31
Figure 9: Averages of NASA-TLX scores vs. task type and noise level ...................................... 32
Figure 10: Task type, noise level, and task type*noise level interaction connecting letters reports
for NASA-TLX scores .................................................................................................................. 34
Figure 11: Correlation scatterplot with ellipse of agreeableness vs. office noise mean pupil
diameter......................................................................................................................................... 45
Figure 12: Correlation scatterplot with ellipse of agreeableness vs. data entry mean pupil
diameter......................................................................................................................................... 46
Figure 13: Correlation scatterplot with ellipse of neuroticism vs. office noise MWL (NASA-
TLX) ............................................................................................................................................. 46
Figure 14: Correlation scatterplot with ellipse of agreeableness vs. anomaly detection mean pupil
diameter......................................................................................................................................... 47
viii
Figure 15: Correlation scatterplot with ellipse of agreeableness vs. no noise mean pupil diameter
....................................................................................................................................................... 47
Figure 16: Graphical illustration of all mental workload measures for task type ......................... 52
Figure 17: Graphical illustration of all mental workload measures for noise level ...................... 53
ix
List of Tables
Table 1: HR and HRV parameters with expected changes with increased mental workload ....... 10
Table 2: Independent and Dependent Variable Lists .................................................................... 17
Table 3: ANOVA for task performance ........................................................................................ 29
Table 4: ANOVA for task duration .............................................................................................. 31
Table 5: ANOVA for NASA-TLX mental workload scores ........................................................ 33
Table 6: ANOVA for LF/HF ratio ................................................................................................ 35
Table 7: ANOVA for mean HRV (not standardized) ................................................................... 36
Table 8: ANOVA for standardized mean HRV ............................................................................ 37
Table 9: ANOVA for standard deviation HRV ............................................................................ 38
Table 10: ANOVA for root mean squared differences of successive R-R intervals (RMSSD) ... 39
Table 11: ANOVA for mean difference in pupil diameter ........................................................... 40
Table 12: ANOVA for pupil diameter standard deviation............................................................ 41
Table 13: Table of f-ratios/t-ratios and p values for data entry task's fixation rate, duration, and
counts ............................................................................................................................................ 42
Table 14: Table of f-ratios/t-ratios and p values for anomaly detection task's fixation rate,
duration, and counts ...................................................................................................................... 42
Table 15: Summary table of physiological parameters when mental workload increases ........... 43
Table 16: Correlation coefficient, p-values, and variables for all correlations that showed
significance ................................................................................................................................... 44
Table 17: Top five correlations from Table 15 ............................................................................. 55
x
ACKNOWLEDGEMENTS
I would like to thank my graduate advisor Dr. Mary Fendley for her patience, dedication,
kindness, and support throughout my college experience. I would like to thank Dr. Ciarallo for
his assistance and willingness to answer all of my questions no matter how long it took. Another
thank you to Dr. Sherwood for also giving me input when I was hitting some dead ends. Special
thanks to Wright State’s Biomedical, Industrial, and Human Factors Engineering Department for
their hard work and care for students. I would also like to thank Dr. Joe Tritschler for hiring me
on, allowing me to focus on school rather than where my funding is coming from. I want to also
mention my great appreciation to my research colleagues Alex Dominic, Noel Fleeman, and Josh
Pilcher for their willingness to help me out in the last months of this research.
1
1.0 Introduction
1.1 Background
Human beings are able to control several different aspects of their work environment;
however, there are some aspects that cannot be changed and must be tolerated. Whether this
includes temperature, atmosphere, lighting, mood, or sound level, these aspects have an
influence on the work environment and, in turn, the humans. Noise is an aspect that is present in
many work environments and is easily controlled to benefit the persons working in the
environment. For example, some individuals often choose a quiet space away from people, while
others choose a space with lots of people and plentiful noise. Additionally, while some
individuals may play soft music without lyrics to focus on their work, others may blast lyrical
melodies until their work is completed.
So, how does one know whether the environmental sound around an individual is
supporting accurate task completion, in an effortless timely manner? Multiple studies have found
that task performance deteriorates when noise is produced in the background (Jahncke,
Björkeholm, Marsh, Odelius, & Sörqvist, 2016; Jerison, 1959; Levy-Leboyer, 1989; Nassiri et
al., 2013; Weistein, 1974). Other studies have shown that not only task performance is impacted.
The Reaction time and the mental load on the human during the task is also impacted (Becker,
Warm, Dember, & Hancock 1995; Lahtela, Niemi, Kuusela, & Hypén, 1986; Ljungberg &
Neely, 2007; Nassiri et al., 2013; Tafalla, & Evans, 1997).
Studies have found that personality may have an impact on the quality of a subject’s
performance when there is noise in the background (Belojevic, Jakovljevic, & Slepcevic, 2003;
Dobbs, Furnham, & McClelland, 2011; Furnham, & Strbac, 2002; Jafari & Kazempour, 2013,
Kou, Furnham, McClelland, & Furnham, 2018). A few other studies have looked at the effect of
2
noise on a human based on the ambient, environmental sounds experienced within the human’s
daily life (Dockrell, Shield, & Dockrell, 2008; Lercher, Evans, & Meis, 2003; Pujol, et al.,
2014).
This research study utilized different types of tasks to examine whether performance,
mental workload, or duration of task are impacted when a subject was exposed to three different
background noise levels. The tasks used in this study were designed to simulate those specific to
professionals in a medical field while the accompanying ambient noise was based on sounds
heard in a typical medical environment (e.g. people talking, copier running, typing, and phone
ringing). Correlation tests were then used to study whether the noise and task type’s influences
relate back to the personality and noise tolerance of the individual. In addition, correlation were
examined to look for a relationship between the effect of noise and task types versus the
personality and lifestyle of the subject.
1.2 Research Objective
The purpose of this study was to assess the influence of noise on human subject task
performance, more specifically: the time it took to complete the tasks, the performance of the
subject, and the mental workload of the subject. Noise tolerance and personality correlation
analyses were taken into consideration to examine the effects of the individual’s personality on
the dependent variables when noise was introduced.
2.0 Literature Review
2.1 Task and Multiple Resource Theory
3
This study incorporates three types of tasks (data entry, anomaly detection, and
mathematical arithmetic) that have been commonly used to study human performance and
mental workload (Cail & Aptel, 2003; Church, 2015; Colligan, Potts, Finn, & Sinkin, 2015;
Dimitrakopoulos et al. 2017; Gabbard, 2017; Galy & Mélan, 2015; Khan & Rizvi, 2010; Kotani,
Takamasu, & Tachibana, 2007; Nickels, 2014; Peng, He, Ji, Wang, & Yang, 2006; Piasecki,
2016). Each of these tasks were designed to simulate actual healthcare practices performed by
different medical personnel. The first task, data entry, relates to the job that a nurse must
complete to catalogue patient information into Electronic Health Records (EHR) (graphical user
interface symbolizing EHR interface found in Appendix B). In fact, Colligan, Potts, Finn, and
Sinkin (2015) analyzed mental workload of pediatric nurses during a data entry task of filling out
EHRs in their actual work environment. The study examined the reaction of the nurses when
switching to the new EHRs, and found that mental workload increased only during the initial
switch to the new EHR.
Both Gabbard (2017) and Piasecki (2016) utilized an anomaly detection task to
investigate mental workload and overall performance of human subjects. Gabbard’s anomalies
were set in a video; while Piasecki had a stagnant scene for the anomalies, the anomalies
sometimes flashed and moved within the still scene. These types of anomalies are in contrast to
this research study’s focus, since both the anomalies and the setting were static. Sets of x-rays
were used in a stagnant image set for the subject to pick out abnormalities (images found in
Appendix B). These stagnant scenarios were appropriate, since a radiologist’s work radiologist’s
profession is far more complex than the task for this experiment and requires years of training ;
thus, the task in this research only attempts to simulate a limited portion of the full
responsibilities of a radiologist in interpreting images.
4
The complexity of the actual medical professional’s job also effected the choice of the
mathematical/arithmetic task. An arithmetic task was simulated with counting and categorizing
medical pills, a task that pharmacists are familiar with and one that this study has dubbed the
“pharmaceutical task” (set-up and problems found in Appendix B). Filling prescriptions require
years of pharmaceutical training and is not a task that just anyone can pick up and perform;
therefore, this study simplified the task utilizing basic math, colorful beads, and a manual pill
counter. Arithmetic tasks are a common means of testing mental workload (Dimitrakopoulos et
al. 2017; Galy & Mélan, 2015; Kotani, Takamasu, & Tachibana, 2007; Peng, He, Ji, Wang, &
Yang, 2006). In this study, mathematical arithmetic’s role in mental workload studies and the
way subjects use mental mathematics elicited interesting performance scores.
Some tasks require more resources than others. Multiple Resource Theory (MRT) states
that there are limitations on what an individual can do all at one time, dependent on four factors,
based on the amount of resources that a certain task demands (Basil, 1994; Wickens, 2002;
Wickens, 2008). These factors are processing stages, perceptual modalities, visual channels, and
processing codes. Each factor has two dimensions, and each dimension has two discrete levels.
While all the tasks in this study incorporate visual modalities, the anomaly detection task relies
heavily on the ability to visualize and interpret the differences among x-rays. The mathematical
arithmetic task weighs more heavily on working memory in terms of processing the information
given. However, the data entry task requires both verbal and auditory resources. One of the
resources, auditory, is being disrupted with different noise levels within this study. This theory
drove the choice of different types of tasks in the design of this experiment to study the effect of
noise on performance, task duration, and mental workload.
5
2.2 Noise Interruption
Several studies have shown that introducing noise to an individual during a task has
measurable effects on workload and task performance (Becker, Warm, Dember, & Hancock,
1995; Gabbard, 2017; Hygge, & Knez, 2001; Jerison, 1959; Levy-Leboyer, 1989; Szalma &
Hancock, 2011; Tafalla & Evans, 1997). In addition, a number of these studies characterize noise
in different ways: low frequency hum, music, office noise, background speech (Dobbs et al.,
2011; Furnham, & Strbac, 2002; Jafari & Kazempour, 2013; Jahncke, Björkeholm, Marsh,
Odelius, & Sörqvist, 2016). However, they all have the commonality that noise is considered to
be an auditory interruption to the main work of the subjects in the experiment.
MRT is based on the ideaof multiple separate resources (verbal, auditory, visual,
perceptual, cognitive, and spatial) with a limited capacity. Each task presented to an individual
is allocated to a specific resource (Basil, 1994; Rubio et al., 2004; Horrey and Wickens, 2006;
Wickens & Wickens, 2008). Though these resources are separate from each other, there can be
interference or “resource competition” when two or more resources are occupied simultaneously
in a subject (Horrey and Wickens, 2006). This theory explains why human performance during a
task focusing on one source (i.e. visual) can suffer when another resource (i.e. auditory) is
causing interference.
A study at the University of Cincinnati had students perform a detection task under no
noise, low noise, and high noise conditions (Becker et al., 1995). Not only did their detection
performance decrease with noise, but their perceived mental workload increased. Another study
found that noise appeared to have the same effect (Tafalla & Evans, 1997). Given an arithmetic
task with two levels of complexity, the study demonstrated that noise increased heart rate under
6
high effort conditions and thus high workload conditions. When noise was present, the reaction
time slowed and effort was low.
A more recent study analyzed the performance of 40 male subjects during noise
interruption (Nassiri et al., 2013). The task involved the use of hand tools while testing
steadiness and dexterity. The results showed that intermediate noise worsened the subjective
work environment of the participants, but treble noise reduced the subject’s performance.
Another study found similar results when office noise was in the background during task
performance of 30 students (Jahncke et al., 2016). In this case, each participant experienced
office noise in the background during their task. Some participants were split into another
experimental group that wore headphones with nature sounds to mask the office noises. The
students who performed the tasks without masking had lower performance compared to when the
noise was masked.
This research study uses headphones to play the noise. These headphones were worn for
all three noise levels. These noise levels were office noise, white noise, and no noise. Varying
the sound playing in the background with these levels will determine whether one noise level has
a stronger effect than another.
2.3 Mental Workload Analysis
2.3.1 NASA TLX
The NASA Task Load Index (NASA-TLX) is a validated source for measuring subjective
mental workload in studies concerned with the cognitive load on subjects (Hart, 2006; Francisco
Ruiz-Rabelo et al., 2015; Gerhard & de Winter, 2015; Hu, Lu, Tan, & Lomanto, 2016; Liang,
Rau, Tsai, & Chen, 2014; Marquart, Cabrall, & de Winter, 2015; Rubio, Díaz, Martín, & Puente,
7
2004; Sönmez, Oğuz, Kutlu, & Yıldırım, 2017). The dimensions that the NASA-TLX assess are
mental demand, physical demand, temporal demand, performance, effort, and frustration. Each
dimension is rated by subjects on a scale out of 20, and each rating can either be weighted to
obtain a global score or taken as a sum of all six dimensions together to obtain a raw score
(Francisco Ruiz-Rabelo et al, 2015; Rubio et al. 2004). There are other mental workload
questionnaires that exist, such as the Subjective Workload Assessment Technique (SWAT) and
Workload Profile (WP). NASA-TLX was chosen for this study because of its better range for a
rating scale, the number of dimensions that directly relate to the study (mental workload,
frustration, temporal demand, and effort), and the lack of time pressure ratings, which was not
necessary for the study (Rubio et al., 2004).
2.3.2 Eye-tracking
Eye tracking parameters, such as blinks, fixations, saccades, and pupil dilation, are a
relevant way to track mental workload of a subject (Cardona & Quevedo, 2014; Gao, Wang, Li,
Dong, & Song, 2013; Gerhard & Joost, 2015; He, Wang, Gao, & Chen, 2012; Holmqvist et al.,
2011; Marquart, Cabrall, & de Winter, 2015; Tokuda, Obinata, Palmer, & Chaparro, 2011; Van
Orden, Limbert, Makeig, & Jung, 2001). Marquart, Cabrall, and de Winter (2015) utilized an
arithmetic task and viewed the changes in the pupil diameter. The results showed that the mean
pupil diameter and the change in pupil diameter correlated with the difficulty of the arithmetic
problem that was presented to the subject. While He et al. (2012) also investigated the pupil size
as a means of mental workload, they also examined fixation duration. While fixation duration
decreased with increased time pressure of the task for smaller time pressures, it increased when
the pressure began to overload the subject mentally. As in previous studies, He et al. (2012)
showed that pupil size increases with mental workload, as well as time pressure. Gao, Wang, Li,
8
Dong, and Song (2013) compared seven different eye-tracking measures and found that blink
rate was sensitive to overall task complexity and blink duration increased over the task period.
However, Faure, Lobjois, and Benguigui (2016) attempted to use blink rate as a means of
quantifying mental workload and found that there was not a significant correlation.
Cardona and Quevedo (2014) found that blink rate did not vary significantly across
complexity levels; however, large amplitude saccades (i.e. the angular distance the eye moves)
accompanied with a blink was related to high cognitive demands. The main findings of Tokuda,
Obinata, Palmer, and Chaparro (2011) showed that saccadic intrusions (SI) were regularly
observed during higher mental workload. SI is a type of eye movement that is jerky and quick.
Compared to a micro-saccade (also a quick, jerky eye movement) SI has a larger amplitude and
does not usually change gaze direction. Pupil diameter is also related to mental workload;
however, SI was still a better indicator of mental workload for this study (Tokuda, Obinata,
Palmer, & Chaparro, 2011).
Van Orden, Limbert, Makeig, and Jung (2001) found that different eye-tracking
parameters distinguished mental workload. They used target density in a mock air warfare task to
vary the task complexity, thus changing the mental workload of the subjects that performed the
tasks. There was a decrease in blink duration and frequency when the target density (complexity)
increased. The opposite happened with fixation frequency; more fixations occurred when the
target density increased.
This research study utilized several eye-tracking parameters. Pupil diameter mean was
compared to a control due to the different pupil sizes of the subjects. Pupil diameter standard
9
deviation for the subjects was also calculated. Automatic gaze mapping was utilized on two
tasks to assess the fixation rates, fixation durations, and fixation counts during those tasks.
2.3.3 Heart Rate Variability
Heart rate variability can be used as a physiological measure of mental workload in
several ways. There are several HRV parameters that can be calculated by using the R-R
interval. Figure 1 shows an illustration of an R-R interval.
Figure 1: Illustration of R-R interval in EKG signal
Mansikka, Virtanen, Harris, and Simola (2016) provide a table that helps in interpreting
heart rate variability and how it changes with pilot mental workload (PMWL). Though the tasks
presented in this study are not the same as those required of a pilot, they are tasks that may cause
increased cognitive load. Thus, the PMWL presented in Table 1 created by Mansikka et al.
(2016) can relate to this study as well. Table 1 summarizes what occurs to the HRV measures
with this study with increased mental workload. The parameters found in the table can be
calculated using the R-R intervals (also referred to as normal to normal or N-N intervals).
R-R interval
10
Table 1: HR and HRV parameters with expected changes with increased mental workload
Mansikka, Virtanen, Harris, and Simola (2016) were able to differentiate autonomous
nervous system response variation between the task segments, rather than just between rest and
trial conditions. While there were significant HRV/HR differences between segments of the
tasks, there were no significant differences in performance. This is a strange phenomenon
considering that other studies found performance to decrease when mental workload increases
(Hu et al., 2016; Nickels, 2014; Prabhu, Smith, Yurko, Acker, & Stefanidis, 2010).
Analysis of HRV defines two different categories of parameters: time domain measures
and frequency domain analysis. LF, HF, and LF/HF all assess variance over a longer period of
time in terms of frequency. RMSSD, SDNN, and MEANRR utilize the normal to normal beat
intervals (time between heart beats) (Heine, Lenis, Reichensperger, Beran, Doessel, & Deml,
2017; Mansikka et al., 2016; Sugimoto, Kitamura, Murai, Wang, and Wang, 2016). These time
and frequency domain measures have been analyzed in studies concerned with mental workload.
One study examined drivers’ mental workload by using a series of R-R intervals gathered from
the subjects and analyzed them based on rhythmical or morphological features (based on the
quantifying form of the ECG waves) (Heine et al., 2017). Unfortunately, none of the features
could distinguish the different levels of mental workload. Sugimoto, Kitamura, Murai, Wang,
Measure Description Expected Change
MEANRR The mean of RR intervals Decrease
SMEANRRThe mean compared to control
"resting" heart rateDecrease
SDRR The standard deviation of RR intervals Decrease
RMSSD
The square root of the mean squared
differences between successive RR
intervals
Increase
LF/HF
The ratio between the power of low
frequency (LF) and high frequency (HF)
components of HRV
Increase
11
and Wang (2016) not only used the time domain measures of R-R intervals, but the low/high
frequency (LF/HF) components of the ECG as well. The study showed an increase in LF/HF,
thus an increase in mental workload, through specific events during the study.
The following study gathered R-R intervals and used them to obtain frequency domain
and time domain measures. The frequency domain parameters taken from the R-R interval data
was LF/ HF. The ratio of LF/HF is assumed to show a shift in dominance from the
parasympathetic to the sympathetic (Billman, 2013). The ratio of LF/HF has been shown to
increase when mental workload increases, which correlates with this shift to the sympathetic
nervous system. The high frequency range is 0.15 to 0.4 hertz, while low frequency range is 0.04
to 0.15. These frequency components are partitioned from the total variance of the continuous
series of beats (Billman, 2013; Mansikka, et al., 2016). Because the time interval the data was
collected from was short, statistical significance was not expected from the frequency domain
parameter; thus, only the LF/HF ratio was used in the analysis. The time domain parameters that
were taken from the data were the mean of the normal to normal beats, the standard deviation of
the normal to normal beats, and the square root of the mean squared difference (RMSSD)
between normal to normal beats. Because the mean is not normalized, and humans tend to have
different resting heart rates, another parameter, standardized mean R-R intervals, will be
analyzed as well to counteract this phenomenon in this study. The standard deviation examines
the change in the R-R intervals of each subject, so it is not necessary to compare this value to a
control. This is also holds true for the RMSSD value (1).
𝑅𝑀𝑆𝑆𝐷 = √1
𝑛−1∑ (𝑅𝑅𝑖+1 − 𝑅𝑅𝑖)2𝑛−1
𝑖=1 (1)
12
For equating the RMSSD value, the initial normal to normal or R-R interval (𝑅𝑅𝑖) is
subtracted from the next R-R interval(𝑅𝑅𝑖+1). This difference is squared, while the next two
intervals are taken and squared and so on until the end of the specified time interval. After this,
the number of sampled R-R intervals (n) in this time interval are considered via division before
square rooting (Vollmer, 2015). The mean and standard deviation of the R-R intervals are
suspected to decrease whereas the RMSSD is suspected to increase (Mansikka et al., 2016). A
study by Guo, Tian, Tan, Zhao, and Li (2016) agreed with the RMSSD value increasing with
mental workload, but the standard deviation of R-R intervals increased. Yet, another study
conducted by Arnrich, Cinaz, Arnrich, La Marca, & Troester (2011) resulted in a decrease
RMSSD value, disagreeing with Table 1. Research is inconclusive and more is necessary to
definitively prove that a certain change in a physiological parameter like RMSSD or standard
deviation of R-R intervals correlates with a change in mental workload.
2.4 Personality
Personality tests are widely used in psychology to assess the mindset of humans. If
features of personality interact with environmental variables to affect performance, personality
tests can be further used to ultimately customize an individual’s work environment.
Since there are a myriad of personality tests to choose from this study conducted a review
to determine the most appropriate test for the participants of this study. The tests that were
reviewed include the Myers-Briggs, Eysenck Personality Questionnaire, and Big Five Inventory.
Myers-Briggs is among the most common personality tests (Pittenger, 1993; Gerras &
Wong, 2016, Cooper, McCord, Campbell, 2017; Boonghee Neelankavil, de Guzman, & Lim,
2013; Pittenger, 2005, Brotherton, 2012). Every year, millions of copies are distributed to
13
schools, churches, the workplace, and counseling centers (Pittenger, 1993). Many users have
accessed the true form of the Myers-Briggs while others have found an online version that was
not a true copy of the test. Myers-Briggs personalities are based on typologies, or distinct
groups of people that the user taking the test can fit into, such as ISTJ (Introverted, Sensing,
Thinking, Judging) or ENFP (Extroverted, Intuition, Feeling, Perceiving). The assignment of
people into one of 16 groups is restrictive because the idea of the user fitting into more than one
group is not accounted for (Myers, 2016; Pittenger, 2005). While a person may fall into an ISFP
group, their other personality traits that are not covered by the test may closely resemble one of
the other fifteen groups. The Myers-Briggs also will place someone into one group, even though
one of their scores barely allows them into that group because they were on the edge or middle
percentile. If a user takes the test and scores 1 percent more for introversion than extroversion,
they are placed into a group known as “Introverts.” The scale scores for introvert vs. extrovert or
any of the three other personality type pairs must be sufficiently large to make a definite
distinction and place an individual into a group (Pittenger, 2005).
Another personality test that is widely used is the Eysenck Personality Questionnaire
(EPQ). Several studies that have tested a noisy and quiet work environment have provided this
test to their participants (Furnham, & Strbac, 2002; Belojevic, Jakovljevic, & Slepcevic, 2003;
Dobbs et al., 2017; Jafari & Kazempour, 2013). The noisy backgrounds common in the studies
that incorporated the EPQ ranged from a “white noise” effect with a low frequency hum to
music. These studies found that under noisy conditions, extroverts performed better and/or faster
on one or more tasks, where introverts did not improve or even decreased in performance levels
compared to when tasks were performed under quiet conditions.
14
Over the span of 50 years, the EPQ has changed dramatically, with the early form called
Maudsley Medical Questionnaire having 40 items and the Revised Eysenck Personality
Questionnaire (EPQR) in 1985 containing 100 items (Francis, Lewis, & Ziebertz, 2006; Gentry,
Wakefield, & Friedman, 1985). While a shorter form was devised in 1985 (EPQR-S), the 48-
question test requires further research to test the validity (Francis et al., 2006). The EPQ test
surveys an individual’s extroversion, neuroticism, and psychoticism while also producing a “lie
score”, that tells the experimenter whether the person taking the test is lying to make themselves
look good (Gentry et al., 1985).
The last personality test reviewed was the Big Five Inventory. Created in the 1980s, the
Big Five Inventory is composed of short phrases in a 44-item inventory, measured 5 different
dimensions, and takes 5 minutes or less to finish (Rammstedt & John, 2007). Multiple studies
have validated the Big Five Inventory’s use in determining job performance (Aarde, Meiring, &
Wiernik, 2017; Rodriques & Rebelo, 2013; Judge, Rodell, Klinger, & Simon, 2013; Alessandri
& Veccione, 2012). Despite its short test time, the Big Five Inventory has proven to be a reliable
and valid survey around the world (Alansari, 2016; Aterberry, Martens, Cadigan, & Rohrer,
2014; Lovik, Verbeke, & Molenberghs, 2017; Reyes Zamorano, Álvarez Carrillo, Peredo Silva,
Sandoval, & Rebolledo Pastrana, 2014) One study tested the validity of the Big Five using 685
undergraduate students from Kuwait. The study’s primary purpose was to test the reliability and
validity of the Arabic translation of the Big Five (Alansari, 2016). A Belgium paper analyzed a
study that used a Dutch version of the Big Five (Lovik et al., 2017). The study utilized a Flemish
population sample where nearly 10,000 surveys were collected. The analysis of the original
study validated the imposed five-factor structure of the Dutch version of the personality test.
Another study tested participants from a Midwestern university in the United States and found
15
that the reliability score of the Big Five Inventory was acceptable (Aterberry et al., 2014). In
Mexico City, a population of 472 adult male and females took the Big Five Inventory (Reyes
Zamorano et al., 2014). The large sample size was used to determine the reliability of the test,
which was proven in the study using a specific procedure that gave them a reliability score.
Taken together, all of these recent studies found the Big Five Inventory to be reliable, even as it
travelled across borders.
The Myers-Briggs is often passed around to individuals within companies, school, and
churches to assess personality while others take a free yet potentially inaccurate version online.
Because there is the chance that answers will be skewed due to participants already taking a form
of the test before and being placed into a particular typology group, the widespread Myers-
Briggs was not chosen for this study. The EPQ is a test with a lengthy read time, but only three
personality dimensions analyzed. While the three facets in this test do breakdown further, the
three aspects are not favorable for a study that is looking for the effect of not only noise, but
stress when performing a task. Because there are three tasks in the study, each being performed
more than once, a shorter test than the EPQ was desired (Francis et al., 2006). Thus, a personality
test that was quicker, easier, and had more facets that supported the study was preferred. The
Big Five Inventory, as stated previously, takes only a few minutes to administer to a user. The
five dimensions presented in the Big Five were believed to best correlate to the test variables
presented in this study. These items include extroversion vs. introversion, openness vs.
closedness to experience, emotional stability vs. neuroticism, agreeableness vs disagreeableness,
and conscientiousness vs. lack of direction (John, Naumann, & Soto, 2008). In addition to this,
the Big Five Inventory has been validated across different cultures and ethnicities which is an
16
important measure for this study since many of the participants come from a university with a
diverse population of students.
3.0 Methods
3.1 Experimental Design
The goal of this study is to understand whether task performance, mental workload, and
duration of different types of tasks are impacted by noise. The study was a 3 x 3 experimental
design with no repetition of any factor level. Thus, this experiment was neither a within or
between subject design. This type of design allowed for all the factor levels to be experienced
once by every subject without the subject potentially getting used to a level that they have
already experienced. Further tables and illustrations of the experimental design and combination
distribution can be found in Appendix A. The independent variables are task (data entry,
anomaly detection, and mathematical arithmetic) and noise type (white, office, and no noise).
The dependent variables are performance, duration, and mental workload (subjective and
physiological measures).
17
Table 2: Independent and Dependent Variable Lists
3.2 Participants
The experiment included 60 participants, 29 females, and 31 males, recruited from
Wright State University. Ages of participants range from 18-31 (M=22.7, SD=2.1). The subjects
were assigned the all three task types and the all three sounds, but the combination of the sound
condition and task types were determined randomly. The order for each task-noise pairing, total
of nine pairings, was randomized using Microsoft Office Excel 2016. This study was approved
by the Wright State University’s Institutional Review Board. Participants were not monetarily
compensated for their participation, but those who were eligible for extra credit in their classes
for participating in a study were able to use their participation for this credit.
Independent Variable
Levels of
Independent
Variable
ArithmeticPerformance
Mean/Standardized
Mean R-R Intervals
Anomaly DetectionTask Duration
Standard Deviation R-
R Intervals
Data Entry Subjective Mental
Workload (NASA-TLX)
RMSSD of R-R
intervals
Pupil Dilation Mean LF/HF Ratio
Office Noise Pupil Dilation Standard
Deviation Fixation Rate
White NoiseFixation Counts Fixation Duration
No Noise
Dependent Variables
Task Type
Noise Level
18
3.3 Stimuli and Apparatus
The experiment was conducted in the Neuroscience Engineering Collaboration building
Lab 431 where both temperature and lighting were held constant. Prior to the study, a Well at
Walgreens Deluxe Arm Blood Pressure Monitor was used to collect blood pressure and pulse
rate data. Additionally, a noise tolerance questionnaire and the Big Five Questionnaire (found in
Appendix B) were given to the subject at the beginning of the experiment. The NASA-TLX was
prescribed after each task. During the actual experiment, a Biopac Student Lab MP36 Data
Acquisition Unit was used to collect EKG data. Tobii Pro Glasses 2 were placed on each
participant along with headphones. The computer, used for the data entry task, had a monitor that
measured 15 inches diagonal, while the GUI itself on the computer screen measured 2 by 6
inches. A Grafco 5709 manual pill counter was used for the mathematical arithmetic task, while
a Restar 2.4 GHz Laser Presenter was used by the subject to circle anomalies in the anomaly
detection task.
Noise Level. ATH-M40x Professional Monitor Headphones were kept on the participant
throughout the experiment regardless of the sound playing. The noise was placed at a safe level,
approximately 67 dB (Rabinowitzs, 2000). The control level was simply no noise, the next level
was a crowd of people talking vaguely (representing a white noise effect), and the last noise level
was intermittent office noises: copy machine, typing on a keyboard, phone ringing, and jet flying
overhead.
Task Type. The three tasks were related to the medical field. The data entry task mimics the task
a nurse must do to gather patient information and type it into an EHR on a computer (the
graphical user interface used for this task is displayed in Figure 2). The anomaly detection task
19
used four x-rays of the same body part in each of five sets. In each set, one of the four x-rays
had an anomaly. An example x-ray set with the anomaly circled can be found in Figure 3. The
third task was a mathematical arithmetic task that used a pharmaceutical pill counter, pill bottles,
and beads (to represent pills). A photo of the mathematical arithmetic task set-up and problem set
can be seen in Figure 4.
Figure 2: Graphical user interface for data entry task
Figure 3: Anomaly detection task with anomaly circled
20
Figure 4: Arithmetic task set-up with manual pill counter, beads, and pill bottle
3.4 Procedure
After signing the informed consent, the subject filled out the noise tolerance
questionnaire and the Big Five Inventory (questionnaires found in Appendix B). Once
completed, the subject was asked their age. Then, their blood pressure and pulse rate were taken,
and they were asked whether these values were normal. If blood pressure cuff reading was not
successful after three attempts, the pulse rate was taken manually at the left wrist by the principal
investigator. The subject was then asked to clean the skin on both inner ankles and the inner
wrist of their non-dominant side. Once dry, disposable electrodes were attached so the Biopac
system could collect EKG data. Tobii Pro Glasses 2 wearable eye tracker was placed on the face
of the subject. Headphones were placed over the subject’s ears to be worn throughout the
experiment while the glasses were calibrated, and EKG signal was checked for any problems
with the signal. To ensure that the subject could hear the sound coming from the headphones, the
sound was played and they were asked if they could hear the sound and if it was too loud. If the
sound was too loud for the subject, the noise was turned down until the subject was comfortable
with the sound level.
21
The subject was given instruction on each task and verbally confirmed their
understanding of the task to be performed. The anomaly detection task had two example sets of
x-rays given to the subject for brief training on what sort of anomalies will appear in the five
sets. As the subject performed the three tasks, they were exposed to three different background
noise levels. At the end of each task, a NASA-TLX was completed by the subject based on the
task just performed. The experiment was considered complete once all three tasks were finished
and the corresponding NASA-TLX questionnaire for the last task was filled out. The EKG signal
was saved, the eye-tracking recording stopped and stored, and the subject was permitted to
remove the data collecting equipment and leave.
3.5 Data Analysis
3.5.1 Performance Scoring
Each task was graded on a different scale, but all were converted to percentages for
analysis. The anomaly detection task gave the subject three opportunities to identify the anomaly
correctly on each of the five sets. The anomaly detection task was graded on a 15-point scale,
meaning that each set had 3 points possible associated with the 3 guesses allotted. The subject
started with 15 points, and each wrong answer subtracted a point from their total score. The
lowest score possible of 0/15 meant that the subject did not answer any of the anomalies
correctly, where a 15/15 meant that the subject had no wrong answers.
The data entry task was graded on a 10-point scale, each point associated with one entry
on the GUI. A complete wrong answer or blank entries meant no point was received for that
entry. If an entry was misspelled or pieces of the entry were missing, a half point was taken off.
22
So, a 0/10 meant that all the entries were wrong or empty and a 10/10 meant all the entries were
filled in completely with the correct spelling.
The mathematical arithmetic task was the most complex task to score because of all the
possible mistakes that could occur. So, each correct pill in each bottle was worth one point for a
total of 38 points. A point was received for each bead that was correct. Since there was no way to
interpret the mental arithmetic of the subject during the task, each bottle was scored solely on the
correct bead placement for each problem. Each bottle was associated with one problem for a total
of five problems/bottles.
The three scores associated with the three tasks for each subject were placed in JMP® by
SAS® (2014) for ANOVA and correlational testing.
3.5.2 Duration of task
Each task was timed, starting when the principal investigator verbally told the subject to
start and ending when the subject completed the task. None of the subjects were told that the
tasks were timed, and no time constraint was set in place to avoid undesired mental workload
due to time pressure felt by the subject. The stress of noise on each of the tasks was the desired
factor to be assessed, not time pressure. The duration of each task was placed into JMP® by
SAS® (2014) for ANOVA and correlational tests.
3.5.3 NASA-TLX
The NASA-TLX is based on a 20-point scale. The scales from each NASA-TLX were
summed and the raw score was taken for analysis (Rubio et al. 2004). Because there were three
tasks to grade with their associated noise level, each subject had three different raw subjective
23
mental workload scores. The scores were placed into JMP® by SAS® (2014) for ANOVA and
correlational tests.
3.5.4 Eye-tracking
The files created from the Tobii Eyetracking Controller Software were imported into the
Tobii Analyzer software for analysis. Only the anomaly detection task and the data entry tasks
were given areas of interest (AOI) to analyze the fixation rate, fixation duration, and fixation
counts from the task. An AOI provided a set region within the task’s stimulus for which more
information would like to be gathered, i.e. the fixation parameters (Holmqvist et al., 2011). The
AOIs for this study can be seen in Appendix C.
The task intervals were manually marked, and the gaze data for the AOIs was mapped
automatically using the software. All the tasks were analyzed for pupil diameter. Thus, the
anomaly and data entry task had a separate analysis in JMP® by SAS® (2014). All three tasks
and noise levels were included with the dependent variable pupil diameter based on average
difference from the control, which was recorded when no task was being performed from the
start of the recording to start of the first task.
3.5.5 Heart Rate Variability
Using Python IDLE Version 3.6, Biopac acquisition files were converted into a readable
text file format. A MATLAB (The Mathworks, Inc) code scanned through the file to obtain the
R-R intervals. These intervals were sifted through to obtain the LF/HF, R-R interval means,
standard deviation of R- R intervals, and the root mean squared differences of successive R-R
intervals. There were three total numbers taken for each parameter from each subject to compare
the three tasks and the three noise levels. For control comparison, the first 60 seconds before the
first task started was recorded from most subjects. Not all subjects had a full 60 seconds before
24
the start of their task, and instead, the chunk of time before the task was recorded instead. During
this time, the subject was given instructions. Recording the EKG signal only 60 seconds before
the task ensured that these longer intervals with issues where electrodes may have needed
adjustment were not taken as the control. This control allowed for further comparison between
when the subject was not working a task to when they were for the standardized mean R-R
interval parameter.
3.6 Hypotheses
3.6.1 Performance
This study hypothesized that when noise interrupts the human subject, task performance
will decrease. Based on the hypothesis, the no noise condition should show the best performance
scores on average, and the office noise (deemed as the more disruptive noise condition) should
show the lowest performance averages. As for task conditions, it is predicted that being a simpler
task, the data entry task will show relatively high scores. However, due to the concept of MRT,
this task will have worse performance scores when noise interrupts (the interaction of this task
and noise will show lower scores than other tasks and noise interactions). Data entry requires the
auditory and verbal cognitive channels while the other tasks do not. The noise will interfere with
the auditory channel, which leaves more possibilities for errors.
3.6.2 Duration of the Tasks
The study hypothesized that the duration of the task will increase when noise is played in
the background. The reasoning for this was that when the subject is interrupted, the subject must
recover before returning to the task at hand, a process that will require time. The task that is
predicted to take the most time is the arithmetic task due to the intense working memory and
computation necessary for solving the problems and the motor skills necessary for maneuvering
25
the beads and placing them into the bottles. However, the task type and noise interaction that is
predicted to show the higher task durations, again, the data entry task with office noise due to the
auditory channels being overwhelmed when they are needed for the task.
3.6.3 NASA-TLX Mental Workload
The NASA-TLX subjective workload has different predictions in terms of the task.
Despite MRT, it is hypothesized that the subject will rate the arithmetic task higher because of
the amount of working memory necessary to complete this task. The noise that is predicted to
show the higher raw NASA-TLX scores on average is the office noise because of the disruptive
work environment the sounds create for the subject. As far as interaction, the office noise and
arithmetic task interaction is predicted to show the higher subject mental workload scores
because of the intermittent, non-continuous nature of this auditory work environment and the
mental math the subject must do to complete the task.
3.6.4 Eye-tracking and HRV Parameters
The eye-tracking parameters that were analyzed are pupil diameter for all tasks and
fixation rates, counts, and durations for the anomaly detection and data entry tasks. It is
hypothesized that an increase in pupil diameter will indicate an increase in subject mental
workload (He et al., 2012). This parameter was normalized by taking the difference of it and the
control (time before the first task start). However, the pupil diameter’s standard deviation (not
normalized) will decrease with increased workload (Othman & Romli, 2016). For heart rate
variability, the predictions follow those of Table 1 when mental workload increases: mean and
standard deviation of R-R intervals will decrease, root mean squared differences of successive R-
R intervals will increase, and the LF/HF ratio will increase. If the mean is suspected to decrease
when mental workload increases, then the standardized version of this parameter should do the
26
same if the task’s values are subtracted from the control. The control should have a greater heart
rate mean than when the subject is performing a mentally demanding task. The more demanding,
the lower mean, the less the difference between the control and task mean R-R intervals will be
(some will even be negative). It is predicted that the same tasks, noises, and task type noise
interactions observed as mental overload using subjective measures will also be observed using
physiological measures.
Fixation rates, counts, and durations were analyzed in two separate one-way ANOVAs in order
to ensure that noise was the only factor analyzed. It is hypothesized that the office noise will
create a more mentally taxing work environment for the subject because it is more disruptive.
The no noise condition will show less of an impact on mental workload. For these parameters,
when mental workload increases, the fixation counts will increase while the fixation duration and
rates will decrease (He et al., 2012; Holmquist et al., 2011; and Van Orden et al., 2001).
4.0 Results
From this study’s 60 subjects, the data for 2 subjects were completely excluded due to an
issue with the sound system and a misunderstanding with one of the subject’s arithmetic tasks.
To perform the statistical analysis, all the data was placed into JMP® by SAS® (2014). Two-
way ANOVAs for the two independent variables (task type and noise level) was performed. All
residual plots testing for normality and unequal variances can be found in Appendix D. The
connecting letters reports, created using Tukey’s HSD test, and interaction plots (for
performance, task duration, and NASA-TLX scores) can be found in Appendix E.
All analysis of the physiological measures of mental workload used the same sample size,
but a different sample size than the performance, duration, and NASA-TLX analysis. Subjects
27
were excluded from the analysis if less than 70% gaze data was collected, their EKG signal was
too noisy, or a technical difficulty occurred. After consideration, a total of 10 of the 60 subjects
were removed from this analysis. The eye-tracking and HRV data was also analyzed using
JMP® by SAS® (2014). However, instead of a two-way ANOVA for the AOI parameters
(fixation counts, fixation duration, and fixation rate) a one-way ANOVA was conducted for each
of the two tasks that had AOIs (data entry and anomaly detection). The values provided in this
section with ANOVA as the statistcal analysis are written as (mean ± standard error). The F ratio
and p-value are also given in the text as well as in the ANOVA tables. Both Microsoft Office
Excel 2016 and JMP® by SAS® (2014) were used for the correlation tests of personality and
noise tolerance measures, two factors that could not easily be controlled in the experiment.
4.1 Performance
There was no strong evidence, provided by the residual plots in Appendix D, that the
variances were different for the performance scores among the three tasks. The residual plots that
tested for normality did not show significant evidence to say that the data was not normally
distributed.
28
Figure 5: Average task type and noise level vs. performance (error bars are standard deviation)
Figure 5 shows a graph of the performance scores average of the different task type under
the three noise conditions. The lowest performance appears to be from the anomaly detection
task while the arithmetic task appears to have the higher scores. Statistical analysis agrees with
this trend. The model showed significance in terms of performance [F(8, 165) = 13.22,
p=<.0001]. The task type with the highest performance score average was the arithmetic task
(92.60 ± 1.78) while the lowest was the anomaly detection task (70.02 ± 1.77). The noise with
the highest performance average was white noise (86.82 ± 1.77) and the lowest was office noise
(81.3 ± 1.78). As far as interactions, the highest performance score was the arithmetic task with
white noise (96.18 ± 3.02) and the lowest performance score average is the interaction of
anomaly detection and office noise (64.33 ± 2.95).
29
Table 2 shows the ANOVA table for the performance metric. The task type factor
showed significance in influencing the performance scores (p-value < 0.0001), but the noise
level did not. Figure 6 shows the connecting letters reports for each task level and the
interactions that further reveals the separation of anomaly detection from the rest of the tasks for
performance scores.
Table 3: ANOVA for task performance
Source DF F ratio p-value
Model 8 13.2252 <.0001
Error 165
Total 173
Source DF F ratio p-value
Task Type 2 49.8471 <.0001
Noise 2 1.5199 0.2218
Task Type * Noise 4 1.6065 0.175
Figure 6: Task type, noise level, and task type*noise level interaction connecting letters reports for performance
30
4.2 Duration
According to the residual plots in Appendix D, there was no strong evidence to suggest
that there were unequal variances for the duration times (recorded strictly in minutes, not
seconds) or that the duration data was not normally distributed.
Figure 7: Average of task duration vs. task type and noise level
From Figure 7, the data entry task appears to take the shortest amount of time to complete
and the arithmetic task takes the longest. The statistics show the same conclusion as the averages
presented in the graph. The model showed significance in terms of duration [F(8, 165) = 12.30,
p=<.0001]. The arithmetic task had the longest task duration average (4.46 min ± 0.19) and the
shortest task duration average for task type was data entry (1.89 min ± 0.19). The longest task
duration average for noise level was no noise (3.51 ± 0.19) and the shortest task duration average
31
for noise level was white noise (2.85 ± 0.19). For task type and noise interactions, the longest
task duration average occurs in the interaction of arithmetic task and no noise (4.90 ± 0.32) and
the shortest average duration occurs in the interaction of data entry task and white noise (1.66 ±
0.32).
Table 3 displays the ANOVA table for the task duration times. Figure 8 displays the
differences between the task levels and the connecting letters report and how the no noise and
white noise level are different, but both connected to the disruptive noise level.
Table 4: ANOVA for task duration
Source DF F ratio p-value
Model 8 12.30435 <.0001
Error 165
Total 173
Source DF F ratio p-value
Task Type 2 45.5581 <.0001
Noise 2 2.9149 0.057
Task Type * Noise 4 0.202 0.937
Figure 8: Task type, noise level, and task type*noise level interaction connecting letters reports for task duration
32
4.3 NASA-TLX Mental Workload
According to the residual plots in Appendix D, there was no strong evidence to suggest
that there were unequal variances for the NASA-TLX raw or that the data was not normally
distributed.
Figure 9: Averages of NASA-TLX scores vs. task type and noise level
As with the other graphs that display the averages of the dependent variables, the
conclusion drawn from Figure 9 on NASA-TLX scores match that of the statistical analysis.
Data entry appears to have the lowest scores, meaning the lowest stress felt by the subject.
Anomaly detection task seems to have the higher scores in most cases except for the no noise
condition where arithmetic pulls ahead slightly. The model showed significance in terms of
NASA-TLX raw scores [F(8, 165) = 7.885, p=<.0001]. The task type with the highest average of
33
raw NASA-TLX mental workload scores was anomaly detection (52.47 ± 2.32), the lowest raw
score average task type being data entry (28.92 ± 2.32). As for the noise level, the highest
NASA-TLX scores average was found in tasks with office noise playing in the background
(45.24 ± 2.32) and the lowest when white noise played in the background (40.95 ± 2.32). As for
the interactions, the highest scores on average came from the anomaly detection and office noise
interaction (55.86 ± 3.85) and the lowest coming for the no noise and data entry interaction
(25.16 ± 4.05).
Table 4 shows the ANOVA for the subject mental workload raw scores that came from
summing the scores of the NASA-TLX for each task. Figure 10 shows the connecting letters
reports for task type and noise level. It shows that the data entry task is different than the other
two tasks for mental workload scores, but the mental workload scores did not change when the
noise level changed because the letters are all similar.
Table 5: ANOVA for NASA-TLX mental workload scores
Source DF F ratio p-value
Model 8 7.885192 <.0001
Error 165
Total 173
Source DF F ratio p-value
Task Type 2 27.9313 <.0001
Noise 2 0.6995 0.4983
Task Type * Noise 4 1.391 0.2393
34
Figure 10: Task type, noise level, and task type*noise level interaction connecting letters reports for NASA-TLX scores
4.4 Physiological Mental Workload
As stated before, the physiological mental workload measures were given different
sample sizes due to the issues that occurred with the Tobii Eye-tracking and Biopac systems. The
same two subjects that were removed for the main ANOVA were removed from this analysis.
Three more subjects were removed due to the noisy EKG signals while another subject was
removed because the cord for the EKG recording came un-plugged during the experiment,
unbeknownst to the PI until after the task was finished. Five more subjects were removed
because of the low gaze data score (<70%). Thus, the sample size subjected to analysis was 50
subjects as opposed to the original 60 subjects that were run for the experiment. According to
the residuals plots in Appendix D, there was no significant evidence that the variances for any of
the physiological data were unequal or the data was not normally distributed. Appendix E has all
of the connecting letters reports.
4.4.1 Heart Rate Variability Analysis
As stated earlier, there were four parameters that were analyzed for heart rate variability,
one frequency domain (LF/HF) and three-time domain (mean, standard deviation, RMSSD)
measures.
35
The LF/HF ratio showed significance in the model [F(8,141) =4.0032, p=0.0003].When
the analysis was run, the highest LF/HF value for task type came from the anomaly detection
task (0.914 ± 0.038) and the lowest coming from the data entry task (0.65 ± 0.039). The noise
level with the highest LF/HF ratio was office noise (0.812 ± 0.038) while white noise had the
lowest ratio (0.6957 ± 0.0385). The task and noise type that had the highest LF/HF value
average was anomaly detection with office noise played in the background (0.994 ± 0.066) and
the lowest LF/HF average being when white noise was played in the background of the data
entry task (0.606 ± 0.062). An ANOVA table for this ratio is found in Table 5.
Table 6: ANOVA for LF/HF ratio
Source DF F ratio p-value
Model 8 4.003171 <.0003
Error 141
Total 149
Source DF F ratio p-value
Task Type 2 11.4415 <.0001
Noise 2 3.0275 0.0516
Task Type * Noise 4 0.5664 0.6874
The non-standardized mean did not show significance within the model [F(8,141) =
1.649, p=0.116]. However, the task type did have a statistically significant impact on the non-
standardized mean, despite the difference in resting heart rates [F(2, 141) = 4.7176, p=0.0104].
The task type with the lowest mean R-R intervals on average was data entry (0.694 ± 0.015)
while anomaly detection had the highest (0.755 ± 0.015). The task type that showed the highest
mean R-R intervals was anomaly detection (0.755 ± 0.015) while the lowest was data entry
36
(0.694 ± 0.015). The highest mean R-R interval for noise level was office noise (0.723 ± 0.015)
while the lowest was white noise (0.709 ± 0.015). The interaction with the highest mean R-R
interval average was anomaly detection with office noise (0.770 ± 0.026). The lowest values for
interaction were found in the interaction of data entry with office noise (0.675 ± 0.026). The
standardized mean showed significance in the model [F (8, 141) =3.82, p=0.004]. For the
standardized mean, the highest task type was anomaly detection (0.055 ± 0.01) and the lowest
was data entry (-0.008 ± 0.01). The noise level that had the highest standardized mean was
office noise (0.022 ± 0.01) and the lowest was white noise (0.008 ± 0.01). When anomaly
detection had office noise played in the background, the standardized mean was the highest on
average (0.067 ± 0.017), and arithmetic task with white noise in the background had the lowest
standardized mean (-0.028 ± 0.018). The non-standardized mean and standardized mean
ANOVA tables are presented in Table 6 and 7.
Table 7: ANOVA for mean HRV (not standardized)
Source DF F ratio p-value
Model 8 1.649596 0.116
Error 141
Total 149
Source DF F ratio p-value
Task Type 2 4.7176 0.0104
Noise 2 0.2387 0.7880
Task Type * Noise 4 0.149 0.5831
37
Table 8: ANOVA for standardized mean HRV
Source DF F ratio p-value
Model 8 3.82 0.0004
Error 141
Total 149
Source DF F ratio p-value
Task Type 2 11.673 <0.0001
Noise 2 0.6072 0.5463
Task Type * Noise 4 1.5910 0.1800
Standard deviation of R-R intervals did show significance in the model [F(8,141) = 3.02,
p=0.0036].The task type that showed the lowest standard deviation in R-R intervals on average
was the anomaly detection task (0.07 ± 0.005) while the highest standard deviation came from
the data entry task (0.102 ± 0.005). Office noise, which played in the background for the task,
had the lowest standard deviation (0.08 ± 0.005) and “no noise” in the background had the
highest noise level R-R interval standard deviation average (0.09 ± 0.005). The task type and
noise level pairing, anomaly detection with office noise played in the background, showed the
lowest standard deviation R-R interval average (0.067 ± 0.009). The highest average pairing for
this HRV parameter was the “no noise” level played during the data entry task (0.121 ± 0.010).
Table 8 displays the standard deviation HRV’s ANOVA values.
38
Table 9: ANOVA for standard deviation HRV
Source DF F ratio p-value
Model 8 3.022596 0.0036
Error 141
Total 149
Source DF F ratio p-value
Task Type 2 9.5674 0.0001
Noise 2 0.7837 0.4587
Task Type * Noise 4 1.322 0.2647
RMSSD of the successive R-R intervals showed significance in the model as well,
[F(8,141) = 4.19, p=0.0002]. Thus, the highest RMSSD for a task type was data entry (0.117 ±
0.007) versus the lowest RMSSD coming from anomaly detection (0.0634 ± 0.007). The noise
condition that had the highest RMSSD value on average was no noise (0.094 ± 0.007) and the
lowest RMSSD value was office noise (0.081 ± 0.007). The highest RMSSD from interaction
was data entry with no noise (0.137 ± 0.0137) and the lowest at anomaly detection with office
noise (0.053 ± 0.012). Table 9 displays the RMSSD ANOVA values.
39
Table 10: ANOVA for root mean squared differences of successive R-R intervals (RMSSD)
Source DF F ratio p-value
Model 8 4.192103 0.0002
Error 141
Total 149
Source DF F ratio p-value
Task Type 2 14.815 0.0001
Noise 2 0.8665 0.4226
Task Type * Noise 4 0.9065 0.462
4.4.2 Eye-tracking Analysis
All tasks and noise levels had pupil diameter measures, or the widening of the pupil in
diameter, that were also taken before the start of the first task for each participant (max.60
seconds). This allowed for the pupil diameter to be normalized by taking the pupil diameter
average for each task and subtracting it by the control average pupil diameter. The greatest mean
difference in pupil diameter average compared to the control comes from the task type data entry
(0.16 ± 0.076) while the lowest was anomaly detection (-0.121 ± 0.075). The noise with the
greatest mean difference in pupil diameter was no noise (0.041 ± 0.076) while the lowest was
seen during the white noise condition (-0.014 ± 0.076). Table 10 shows the ANOVA values for
mean difference in pupil diameter.
40
Table 11: ANOVA for mean difference in pupil diameter
Source DF F ratio p-value
Model 8 1.3601 0.2192
Error 141
Total 149
Source DF F ratio p-value
Task Type 2 3.5097 0.0325
Noise 2 0.1741 0.8404
Task Type * Noise 4 0.9966 0.4116
The task that had the lowest standard deviation on average was data entry (0.261 ± 0.015)
while the highest standard deviation came from the arithmetic task (0.309 ± 0.015). The noise
that showed the lowest standard deviation in pupil diameter on average was office noise (0.274 ±
0.015) while the highest was no noise (0.293 ± 0.015). The interaction that had the lowest
average in terms of pupil diameter standard deviation was data entry with office noise in the
background (0.249 ± 0.025) while the highest was found during the arithmetic task with no noise
played (0.334 ± 0.024). Table 11 shows the analysis for pupil diameter standard deviation.
41
Table 12: ANOVA for pupil diameter standard deviation
Source DF F ratio p-value
Model 8 1.0550 0.3982
Error 141
Total 149
Source DF F ratio p-value
Task Type 2 2.7887 0.0649
Noise 2 0.4544 0.6358
Task Type * Noise 4 0.3058 0.8737
Fixation parameters, duration, rate, counts, were collected via automatic mapping by the
Tobii Eye-tracking Analyzer Software. Two AOIs were created, one for the data entry task and
one for the anomaly detection task. These AOIs can be found in Appendix C. Table 12 and 13
below show the f-ratios for the one-way ANOVAs done for the 3 noise levels, and the t-ratios for
each of the levels. The p-values for each are presented showing that there was no significant
evidence that noise level influenced any of the fixation parameters for either task based on the
95% CI.
42
Table 13: Table of f-ratios/t-ratios and p values for data entry task's fixation rate, duration, and counts
Data Entry Fixation Parameters Fixation Rate
Source/Level P-value
Model 0.1732
No Noise 0.2076
White Noise 0.0657
Office Noise 0.6
Fixation Duration
Source/Level P-value
Model 0.9185
No Noise 0.688
White Noise 0.7726
Office Noise 0.8876
Fixation Count
Source/Level P-value
Model 0.631
No Noise 0.339
White Noise 0.5869
Office Noise 0.6283
Table 14: Table of f-ratios/t-ratios and p values for anomaly detection task's fixation rate, duration, and counts
Anomaly Fixation Parameters Fixation Rate
Source/Level P-value
Model 0.7633
No Noise 0.4678
White Noise 0.7812
Office Noise 0.6578
Fixation Duration
Source/Level P-value
Model 0.4691
No Noise 0.5427
White Noise 0.5214
Office Noise 0.2211
Fixation Count
Source/Level P-value
Model 0.2163
No Noise 0.158
White Noise 0.8234
Office Noise 0.1089
43
4.4.3 Summary of Physiological Parameters
Because not all of the physiological parameters presented in the results section agree with
one another, a comprehensive table, Table 14, is presented to summarize the findings. The table
shows the physiological measure, the hypothesis, how the measure actually changed (based off
NASA-TLX subjective score), and the task type that the measure placed as the least mentally
straining task and the most based on the hypothesis. Red text notes that what occurred in the
study did not match the hypothesis. The correlation coefficient and p-value resulted from a
Pearson Correlation Test between the NASA-TLX scores and the physiological parameter.
Table 15: Summary table of physiological parameters when mental workload increases
4.5 Correlation Testing
Microsoft Office Excel 2016 was used to determine correlation for as many measures as
possible, by performing nearly 200 correlation tests using the “CORREL” function. The
significant correlation coefficient comparisons were placed into JMP® by SAS® (2014) for
further analysis. Depending on the correlation test being done, the N value is either 58 or 50,
meaning the df (N-2) is either 56 or 48. Those critical values for 50 and 60, the closest df values
in the table to 56 and 48, are between 0.21 and 0.23 for correlation coefficient or above
(Rummel, 1976). The correlation coefficients whose absolute values met the criterion and
showed statistical significance are featured below in Table 15. The top five, lowest p-values are
Physiological Mental
WorkloadHypothesis
What happened
compared to NASA-
TLX results
Task with highest
MWL if hypothesis
true
Task with lowest
MWL if hypothesis
true
Correlation
coefficient with
NASA-TLX
p-value
LF/HF** Increase Increase Anomaly Detection Data Entry 0.0675 0.4118
MEANRR* Decrease Increase Data Entry Anomaly Detection 0.1292 0.115
SMEANRR** Decrease Increase Data Entry Anomaly Detection 0.2023 0.013
SDRR** Decrease Decrease Anomaly Detection Data Entry -0.0765 0.3522
RMSSD** Increase Decrease Data Entry Anomaly Detection -0.0886 0.281
Pupil Diameter (Mean) * Increase Decrease Data Entry Anomaly Detection -0.1689 0.0388
Pupil Diameter (Standard
Deviation)Decrease Increase Data Entry Arithmetic
-0.0089 0.9139
* = Task Type Statistical Significance ** = Task Type and Model Statistical Significance
44
highlighted in green. Correlation coefficient values found using Microsoft Excel can be found in
Appendix F.
Table 16: Correlation coefficient, p-values, and variables for all correlations that showed significance
Variable 1 Variable 2 Correlation Coefficient p-value
Agreeableness Office Noise Mean Pupil Dilation -0.442 0.0013
Agreeableness Data Entry Mean Pupil Dilation -0.4036 0.0037
Agreeableness Anomaly Detection Mean Pupil Dilation -0.3881 0.0054
Agreeableness No Noise Mean Pupil Dilation -0.3766 0.007
Agreeableness Data Entry RMSSD 0.3636 0.0094
Agreeableness Arithmetic Mean Pupil Dilation -0.3601 0.0102
Agreeableness No Noise LF/HF -0.3498 0.0128
Agreeableness Arithmetic Pupi l Di lation Standard Deviation -0.3201 0.0234
Agreeableness Data Entry Standard Deviation HRV 0.3144 0.0262
Agreeableness White Noise Mean Pupil Dilation -0.2835 0.046
Agreeableness No Noise RMSSD 0.2819 0.0474
Agreeableness Noise Tolerance 0.2383 0.0716
Agreeableness No Noise Standard Deviation HRV 0.2494 0.0806
Agreeableness No Noise Mean HRV -0.2421 0.0902
Agreeableness White Noise RMSSD 0.2404 0.0927
Conscientiousness No Noise Mean HRV -0.3552 0.0114
Conscientiousness Data Entry LF/HF -0.3502 0.0127
Conscientiousness Arithmetic Mean HRV -0.3181 0.0244
Conscientiousness No Noise LF/HF -0.2902 0.0409
Conscientiousness Data Entry Mean HRV -0.288 0.0426
Conscientiousness White Noise Mean HRV -0.2618 0.0663
Extroversion Anomaly Detection MWL -0.3174 0.0152
Extroversion Office Noise Duration of Task 0.221 0.0956
Neuroticism Office Noise MWL 0.3664 0.0047
Neuroticism Data Entry RMSSD -0.3462 0.0138
Neuroticism Data Entry Standard Deviation HRV -0.3216 0.0228
Neuroticism Arithmetic MWL 0.2895 0.0275
Neuroticism MWL Overall Average 0.2876 0.0286
Neuroticism White Noise Standard Deviation HRV -0.3026 0.0327
Neuroticism Average Noise (White and Office) MWL 0.2661 0.0435
Neuroticism Anomaly Detection MWL 0.265 0.0444
Neuroticism White Noise RMSSD -0.2619 0.0662
Neuroticism Office Noise Standard Deviation HRV -0.2438 0.0879
Neuroticism No Noise Mean HRV -0.2421 0.0902
Neuroticism Anomaly Detection LF/HF 0.2357 0.0994
45
The personality trait that appears most frequently in Table 14 was agreeableness. The
second personality trait that appeared numerous times was neuroticism. The third most frequent
personality trait was conscientiousness, and the last personality trait that showed significance, yet
least common, was extroversion. Openness and noise tolerance failed to show significance in
terms of correlation tests. The dependent variable data was gathered by task, but had to be sorted
into noise categories, there were two difference sample sizes (different sample size=different df)
for the main ANOVA dependent variables (performance, task duration, and NASA-TLX scores)
and the physiological dependent variables within the correlation tests, The NASA-TLX is
represented in the table and correlation graphs below as simply MWL. Because there were
several correlation tests that showed significance, the top five with the lowest p-values will be
discussed. The smallest p-value was seen in the test between agreeableness personality scores
and office noise mean pupil diameter (r = -0.442; p = 0.0013); the scatterplot of these variables is
shown in Figure 11. The scatterplots with the ellipse illustrates the general correlation direction
between the two variables.
Figure 11: Correlation scatterplot with ellipse of agreeableness vs. office noise mean pupil diameter
46
The next lowest p-value was the agreeableness personality test scores and the data entry
mean pupil diameter (r = -0.4036; p = 0.0037). The third lowest correlation test p-value was
neuroticism personality scores and the office noise MWL, or NASA-TLX raw scores (r =
0.3664; p = 0.0047). These two scatterplots can be found in Figures 12 and 13.
Figure 12: Correlation scatterplot with ellipse of agreeableness vs. data entry mean pupil diameter
Figure 13: Correlation scatterplot with ellipse of neuroticism vs. office noise MWL (NASA-TLX)
47
The fourth lowest p-value amongst the correlation test was agreeableness personality
scores and anomaly detection mean pupil diameter (r = -0.3881; p = 0.0054). The fifth lowest p-
value to be discussed is the test between agreeableness personality scores and no noise mean
pupil diameter (r = -0.3766, p = 0.007). These two scatterplots can be found in Figure 14 and 15.
Figure 14: Correlation scatterplot with ellipse of agreeableness vs. anomaly detection mean pupil diameter
Figure 15: Correlation scatterplot with ellipse of agreeableness vs. no noise mean pupil diameter
48
5.0 Discussion
5.1 Performance, Duration, and NASA-TLX
The main analysis of this study, concerning the impact on performance, task duration,
and mental workload when noise was introduced in the background, had noticeable findings that
both supported and rejected the hypotheses. It was hypothesized that when noise was played in
the background, especially the more disruptive office noises, the subject’s performance would
decrease, while their task duration and mental workload would increase.
There was significant evidence supporting that task type impacted the performance, task
duration, and subjective mental workload. The effect of noise level and task type* noise
interaction was not significant for all three of these dependent variables, so there was no
evidence that noise affected different task types differently. While the lowest performance score
for noise was with office noise, the highest performance scores came from the white noise work
environment. While the office noise was hypothesized to cause a depletion in performance, it is
surprising that the white noise condition showed the higher scores rather than the no noise
condition. It is possible that the subjects who participated are accustomed to working in
environments where people are talking, which was the chosen background sound for the white
noise condition. White noise had the lower NASA-TLX scores, which might also explain the
better performance scores because the subject felt less mentally overloaded. This phenomenon
also explains why the anomaly detection task, which had subjectively higher mental workload
among subjects, had the lowest performance scores on average.
This phenomenon does not explain why the task that was perceived less mentally
straining, data entry, did not have the highest scores. An explanation for this could be that
because the task was perceived too easy, the task’s difficulty was underestimated, resulting in
49
careless errors. The arithmetic task had the highest scores, and this task type fell in the middle as
far as NASA-TLX scores. One explanation for this observation is that too much mental
workload decreases performance, but so does too little mental workload. There may be a
“happy-medium” mental workload caused by the complexity of a task that supports good
performance and is supported by Yerkes-Dodson Law (Wickens, Lee, Liu, & Becker, 2004)
The best performance averages occurred when white noise/arithmetic tasks were
together, and the lowest performance averages occurred when anomaly detection/office noise
were together. This result is consistent with the results for each factor separately. Additionally,
the NASA-TLX scores for task type and noise level interaction that were the highest showed the
lowest performance scores. However, the highest performance score interaction did not
necessarily have the lowest mental workload. On the contrary, the lowest NASA-TLX scores
occurred during the no noise data entry condition. The data entry task required the auditory
cognitive channels of the subject; so having no noise to crowd these channels is consistent with
the perceptions of this condition being “simpler”, resulting in lower NASA-TLX scores.
The duration analysis also rejected and supported the hypotheses. Task duration is one of
two dependent variables be on the margin of statistical significance. It was predicted that when
background noise was present, the task duration would increase. The longest durations were
predicted to be seen in the arithmetic task type and the data entry task with office noise
interaction. The task type results follow the hypothesis, but the results for the arithmetic task
with no noise condition had the longest duration. The noise level that had the longest duration
was no noise, and white noise had the shortest duration on average. The noise results
contradicted the hypothesis that background noise causes longer task duration due to attention
gravitating toward the distracting sounds with corresponding recovery time. It is possible that
50
introducing noise during a task creates a work environment that the subject feels they must move
faster. Having no noise may have relaxed the subject to a point where they took their time
completing the task.
The shortest durations came from the data entry task and data entry task with white noise
interaction. These results make it clear that the data entry task may have just taken a shorter
amount of time to complete. However, it is interesting that both the interaction and noise level
show the white noise condition taking the shortest time on average. White noise showed the
highest performance out of all the noise levels. In addition, white noise displayed the lowest
NASA-TLX scores compared to no noise and office noise. As stated in the beginning, the ideal
work environment supports a culture of high performance, low mental workload, and quick task
completion. It might not be a coincidence that white noise appears to encourage this
environment. Further research should be conducted to confirm that white noise in the
background can create a positive work environment.
5.2 Physiological Mental Workload
All the physiological parameters, except for pupil diameter standard deviation, showed
statistical significance for task type. The LF/HF ratio nearly showed significance for noise level
with a 95% CI. Whether the measures were predicted to decrease or increase with mental
workload, the main hypothesis for these physiological measures were that they would agree with
the subjective mental workload quantified by the NASA-TLX scores.
The mean, standard deviation, and standardized mean values of R-R intervals were
hypothesized to decrease when mental workload increased (Mansikka et al. 2016). Relative to
the NASA-TLX scores, the only parameter that supported the hypothesis was the standard
51
deviation of R-R intervals, which showed the same task, anomaly detection, to have the most
mental workload. If the mean and standardized mean of the R-R intervals decreased with
increased mental workload, this meant that the data entry task had the highest mental workload,
which disagrees with the NASA-TLX. The RMSSD had the same trend. The hypothesis was
that this value would increase along with the mental workload; but if this were true, the data
entry task had the highest mental workload of all the tasks, disagreeing with the NASA-TLX.
The only other HRV parameter that agrees with the NASA-TLX is the LF/HF ratio. LF/HF
ratio, the other dependent variables that nearly showed noise level statistical significance, shows
that the office noise has the highest mental workload and white noise has the lowest.
The pupil dilates when under stress, and this study hypothesized that the difference in
pupil diameter should increase with mental workload (He et al., 2012; Marquart et al., 2015).
The data entry task showed the highest mental workload for pupil dilation, which does not match
the subjective mental workload scores. However, the pupil diameter standard deviation showed
similar results indicating data entry had the highest mental workload when the diameter
decreased, again, disagreeing with the NASA-TLX. Unlike the other physiological measures
that state that either anomaly detection or data entry are the highest or lowest, pupil diameter
standard deviation shows that the arithmetic task has the lowest mental workload.
To better illustrate the contradicting mental workload measures, Figure 16 displays the
trend for each. The NASA-TLX trend is based on the results, and the physiological measures are
based on the results as a function of how the measure should indicate an increase in mental
workload (see Table 14). As stated above, the only two physiological parameters that appear to
agree with the NASA-TLX scores are the SDRR (standard deviation of R-R intervals) and the
LF/HF ratio.
52
Figure 16: Graphical illustration of all mental workload measures for task type
Figure 17 displays mental workload dependent variable trends in terms of noise level.
The only physiological parameter that agrees with the NASA-TLX is the LF/HF ratio. The rest
of the physiological measures were inconsistent in their agreement with which mental workload
level correlates with each noise level. Pupil diameter standard deviation and standard deviation
of R-R intervals show that the office noise has the highest mental workload and no noise has the
lowest, where the RMSSD value results show the opposite. The mean and standardized mean of
R-R intervals show the white noise having the highest mental workload, but pupil diameter mean
shows this noise level as having the lowest mental workload.
Anomaly Detection Data Entry Arithmetic
Low
MW
L
Med
ium
MW
L
Hig
h M
WL
Task Type
NASA-TLX
LF/HF
MEANRR
SMEANRR
SDRR
RMSSD
Pupil Diameter (Mean)
Pupil Diameter (StandardDeviation)
53
Figure 17: Graphical illustration of all mental workload measures for noise level
Despite these findings, there are several explanations for this contradicting data. The
NASA-TLX is considered the gold standard of subjective mental workload, while the
physiological measures are still being validated (Hart, 2006; Francisco Ruiz-Rabelo et al., 2015;
Gerhard & de Winter, 2015; Hu, Lu, Tan, & Lomanto, 2016; Liang, Rau, Tsai, & Chen, 2014;
Marquart, Cabrall, & de Winter, 2015; Rubio, Díaz, Martín, & Puente, 2004; Sönmez, Oğuz,
Kutlu, & Yıldırım, 2017). Previous studies showed mixed results, such as with the RMSSD
value and standard deviation of R-R intervals, with physiological workload measures (Arnrich et
al., 2011; Guo et al., 2016; Mansikka, 2016).
The HRV findings might be skewed because of the noise present in the EKG. The tasks
all required motion: the data entry task required typing; the arithmetic task required a manual pill
counter; and the anomaly detection task required a laser pointer to circle the abnormalities. This
motion created noise in the EKG signal. Because of this noise, the signal was filtered using the
Office Noise White Noise No Noise
Low
MW
L
Med
ium
MW
L
H
igh
MW
L
Noise Level
NASA-TLX
LF/HF
MEANRR
SMEANRR
SDRR
RMSSD
Pupil Diameter (Mean)
Pupil Diameter (StandardDeviation)
54
one of the Biopac software’s digital filters. This data manipulation and/or the leftover noise in
the EKG signal could be a reason why the HRV data contradicted the NASA-TLX.
The pupil diameter findings may also have been skewed. A pupil has multiple jobs, and
some of them, keeping out light in a bright spaces, letting more light in a dark spaces, focusing
on objects different distances away, require the pupil to dilate or constrict (Spector, 1990). Each
task was a different distance from the subject. The data entry task required the subject to look at
a bright computer screen directly in front of them while the anomaly detection task was a couple
feet away. So, even though the lighting was kept constant, the monitor in the data entry task
presented more light than the anomaly detection task while the anomaly detection task was a
greater distance away. These slight differences could account for the pupil mental workload
measures opposing the NASA-TLX.
Fixation duration, counts, and rates were evaluated for the data entry and anomaly
detection tasks. The fixation rate, count, and duration for the anomaly detection task failed to
show any significance. The fixation rate, count, and duration for the data entry task failed to
show any significance as well. Thus, there is not enough evidence to support that noise level
influenced these parameters. This may be due to the large AOIs created for the two tasks that
resulted from the differing heights of the subjects and the distance between the subject and task
varying amongst the experimental sample. The benefit of using a mobile eye-tracking device,
such as the Tobii Pro Glasses 2, is that the subjects can be allowed the mobility similar to a real-
life setting. However, it is highly suggested that the scene be set or adjusted for each participant
so that the eye levels of the participants are comparable to one another to get more accurate
results.
55
5.3 Correlation Tests
Table 15 shows the five correlations with the associated correlation coefficients and p-values.
Table 17: Top five correlations from Table 15
Agreeableness, as four of the five significant correlation tests in Table 15 shows, is
heavily tied with mean pupil diameter. The four conditions where this occurs is the office noise,
and no noise condition, and the data entry and anomaly detection task. All correlation
coefficients to these four tests are negative indicating that when the person is more disagreeable
(has a lower agreeableness score) they have greater mean pupil diameter (pupil dilation). Having
four significant negative correlations between agreeableness scores and the physiological mental
workload parameter pupil diameter (shown as pupil dilation in the table) is interesting. Further
research is necessary, but this trend could make a case for personality tests helping predict a
better work environment in terms of mental workload.
Neuroticism scores were tied to office noise subjective mental workload (MWL)
measured by the NASA-TLX. The correlation coefficient is positive, meaning that when the
neuroticism scores increase, subjective mental workload of the subject also increases during the
office noise condition, despite whatever task is happening during the noise. The office noise
condition was considered the most disruptive noise condition. It is plausible to say that a more
neurotic person, someone with less emotional/mental stability, does not cope well with noise.
Variable 1 Variable 2 Correlation Coefficient p-value
Agreeableness Office Noise Mean Pupil Dilation -0.442 0.0013
Agreeableness Data Entry Mean Pupil Dilation -0.4036 0.0037
Neuroticism Office Noise MWL 0.3664 0.0047
Agreeableness Anomaly Detection Mean Pupil Dilation -0.3881 0.0054
Agreeableness No Noise Mean Pupil Dilation -0.3766 0.007
56
6.0 Conclusion
The goal of the study was to investigate the impact of noise on different tasks performed
by a human subject and determine the effect of noise on performance, mental workload, and time
taken to complete the task. There was significant evidence that task type influenced performance,
duration of task, and mental workload (subjective and physiological). There were two dependent
variables that nearly showed significance in noise level, which was the duration of the task and
the LF/HF ratio, a physiological measure of mental workload. White noise appeared to harbor
the traits of an ideal work environment based on the results: higher performance scores, lower
mental workload scores, and shorter amounts of time spent on tasks. The physiological mental
workload parameters did not necessarily agree with each other or the hypothesis, but most of
them resulted in the same two tasks (anomaly detection and data entry) as either the most and
least mentally taxing respectively. As research continues to move forward with the use of
physiological measures to give real-time indicators of mental workload, it can be better
understood how there are benefits and drawbacks to their use in studies as these. There were
several correlation tests that showed significance; however, agreeableness and neuroticism are
the two personalities that appeared the most in these tests as being significant. It is interesting
that of all the correlation tests performed, the only ones that showed significance were those that
were related to the subjective and physiological mental workload measures. These correlation
tests provide a starting point for analyzing the best work environment for lower mental workload
characterized by the individual’s personality. Additional research must be done to verify the
findings of these correlation tests and further examine the phenomenon of personality’s
correlation with mental workload.
57
7.0 Appendix
7.1 Appendix A: Experimental Design and Combinations Table
Subject Number Anomaly Data Entry Arithmetic
1 No White Office
2 White Office No
3 Office No White
4 Office No White
5 White Office No
6 No White Office
7 No White Office
8 Office No White
9 No Office White
10 White Office No
11 White Office No
12 No White Office
Questionnaires
/ Equipment
Set-up
Arithmetic
Data
Entry
Anomaly
Detection
Office
White
None
Data
Entry
Anomaly
Detection
Arithmetic
None
Office
White
Questionnaires
/ Equipment
Set-up
Questionnaires
/ Equipment
Set-up
Anomaly
Detection
Arithmetic
Data Entry
White
None
Office
Subject 1 Subject 2 Subject 3
Start Start Start
Three examples of subject’s experiment with three tasks with corresponding noises
58
13 Office No White
14 No White Office
15 Office No White
16 White Office No
17 White Office No
18 Office No White
19 White Office No
20 No White Office
21 No Office White
22 Office White No
23 Office White No
24 Office White No
25 Office White No
26 White No Office
27 No Office White
28 No Office White
29 White No Office
30 Office White No
31 Office No White
32 No White Office
33 White Office No
34 White No Office
35 Office No White
36 No White Office
37 Office White No
38 White No Office
39 No Office White
40 White No Office
59
41 No White Office
42 Office No White
43 White Office No
44 White Office No
45 No White Office
46 Office No White
47 No Office White
48 White No Office
49 Office White No
50 Office No White
51 White Office No
52 White No Office
53 Office White No
54 No Office White
55 No Office White
56 White No Office
57 Office White No
58 Office No White
59 No White Office
60 White Office No
Task Type Noise Level
Data Entry White
Data Entry No
Data Entry Office
Anomaly Detection White
Anomaly Detection No
Anomaly Detection Office
Math Arithmetic White
Math Arithmetic No
Math Arithmetic Office
9 Possible Combinations
60
7.2 Appendix B: Questionnaires and Task Problems (with answers)
7.2.1 Noise Tolerance Questionnaire
Participant #: ______
Pre-Questionnaire
Circle the fill-in to the statement that best fits for you. Please circle only one for each statement.
1. I am between 18-65 years old:________.
a. True
b. False
2. I am fluent in the English Language: ________.
a. True
b. False
3. I am not colorblind:__________. (If you could be colorblind, circle False)
a. True
b. False
**If you’ve answered false to any of the above questions, please turn in your pre-questionnaire now
4. Do you have experience with medical imaging or x-rays? ________
a. Yes
b. No
5. Have you taken the Myers-Briggs personality test? __________
a. Yes
b. No
5a. If yes to question 5, what four letters were your result? ___________ (Example: ISTJ)
Reminder or Myers-Briggs:
Introvert/Extrovert (I or E) Sensing/Intuition (S or N)
Thinking/Feeling (T or F) Judging/Perceiving (J or P)
6. I find that I studied the best:
a. In a very quiet place
b. With light music or a small amount of people
c. In a place with lots of people or loud music
61
7. I find that I sleep the best when:
a. It is very quiet
b. With white noise (such as a fan or AC running)
c. When there is a lot of noise outside my window/room
8. I’ve lived over half of my life in:
a. The country/rural (little to no sound)
b. Suburban type area (some sound)
c. City (lots of sound)
9. When I drive/travel in the car, I like to listen to:
a. Little to no music
b. Music mild in volume (can still hold conversation in car)
c. Loud music (must turn down to have a conversation)
10. My ability to cancel out a conversation when I have to concentrate is:
a. Poor
b. Average
c. Excellent
7.2.2 Big Five Inventory (Link)
http://fetzer.org/sites/default/files/images/stories/pdf/selfmeasures/Personality-BigFiveInventory.pdf
7.2.3 NASA-TLX (Link)
https://humansystems.arc.nasa.gov/groups/tlx/downloads/TLXScale.pdf
62
7.2.4 Anomaly Detection Task (with answers)
7.2.5 Data Entry Task (with answers)
63
7.2.6 Mathematical Arithmetic Task (with answers)
7.3 Appendix C: Eye-tracking Illustrations
8 green beads
5 blue beads
Only bead shape and
colors that should be
used
10 blue beads, 4 pink beads, and 2 green beads
2 pink beads and 1 blue bead
1 green bead, 3 pink beads, and 2 blue beads
64
Data Entry Task Area of Interest
Anomaly Detection Task Area of Interest
65
Example data entry gaze plot of subject who only focused on GUI
Example data entry gaze plot of subject who focused on both the keyboard and
GUI
66
Example data entry gaze plot of subject who focused on GUI and Principal
Investigator Example gaze plot for anomaly detection task
67
7.4 Appendix D: Residual Plots
7.4.1 Normal Distribution Checks
68
69
70
71
7.4.2 Residual vs. Predicted Plots
72
73
74
7.5 Appendix E: Connecting Letters Reports and Interaction Plots
7.5.1 Task Performance
75
7.5.2 Task Duration
76
7.5.3 Mental Workload
77
7.5.4 Heart Rate Parameters
LF/HF
Mean R-R Interval
Standard Deviation
78
RMSSD
Standardized Mean HRV
79
7.5.5 Eye-tracking Parameters
Pupil Diameter Mean
Pupil Diameter Standard Deviation
80
7.6 Appendix F: Correlation Tables
Performance Scores by Task vs.
Personality/Noise Tolerance
Scores Noise Tolerance vs.
Anomaly Detection 0.016594
Noise Tolerance vs Data
Entry 0.1715
Noise Tolerance vs.
Arithmetic 0.037862
Agreeableness vs Anomaly
Detection -0.01532
Agreeableness vs. Data
Entry 0.02795
Agreeableness vs.
Arithmetic 0.033015
Conscientiousness vs
Anomaly Detection -0.02843
Conscientiousness vs Data
Entry 0.048
Conscientiousness vs
Arithmetic 0.069696
Neuroticism vs. Anomaly
Detection -0.12451
Neuroticism vs. Data Entry -0.02903
Neuroticism vs. Arithmetic -0.17541
Openness vs Anomaly
Detection 0.094041
Openness vs Data Entry 0.008617
Openness vs Arithmetic 0.03632
Extroversion vs. Anomaly
Detection 0.114438
Extroversion vs Data Entry -0.06425
Extroversion vs. Arithmetic 0.07508
81
NASA-TLX Mental Workload (MWL) scores by
Task vs Personality/Noise Tolerance Scores
Noise Tolerance vs. Anomaly Detection
MWL -0.03498
Noise Tolerance vs Data Entry MWL -0.04264
Noise Tolerance vs. Arithmetic MWL -0.10975
Agreeableness vs Anomaly Detection MWL 0.053231
Agreeableness vs. Data Entry MWL -0.02493
Agreeableness vs. Arithmetic MWL -0.03576
Conscientiousness vs Anomaly Detection
MWL 0.036895
Conscientiousness vs Data Entry MWL -0.008
Conscientiousness vs Arithmetic MWL -0.09948
Neuroticism vs. Anomaly Detection MWL 0.265037
Neuroticism vs. Data Entry MWL 0.128036
Neuroticism vs. Arithmetic MWL 0.272319
Openness vs Anomaly Detection MWL 0.1226
Openness vs Data Entry MWL -0.09011
Openness vs Arithmetic MWL 0.055618
Extrovert vs. Anomaly Detection MWL -0.31736
Extrovert vs Data Entry MWL -0.02714
Extrovert vs. Arithmetic MWL -0.05365
Noise Tolerance vs. MWL Average (over
three tasks) -0.08122
Agreeableness vs MWL Average (over three
tasks) -0.00318
Conscientiousness vs. MWL Average (over
three tasks) -0.03134
Neuroticism vs. MWL Average (over three
tasks) 0.28758
Openness vs. MWL Average (over three
tasks) 0.040067
Extrovert vs. MWL Average (over three
tasks) -0.17175
Personality vs. Noise Tolerance Scores
82
Noise Tolerance vs. Noise Tolerance 1
Agreeableness vs. Noise Tolerance 0.238318
Conscientiousness vs Noise Tolerance 0.134094
Neuroticism vs Noise Tolerance -0.18865
Openness vs Noise Tolerance -0.13839
Extroversion vs. Noise Tolerance 0.11816
Personality/Noise Tolerance Scores vs. Performance
Scores by Noise Level
Extroversion vs. No Noise -0.022748768
Noise Tolerance vs No Noise 0.035836108
Agreeableness vs No Noise -0.000774025
Conscientiousness vs. No Noise 0.11990811
Neuroticism vs. No Noise -0.170906224
Openness vs No Noise -0.057195205
Extroversion vs. White Noise 0.141599354
Noise Tolerance vs White Noise 0.016775334
Agreeableness vs White Noise 0.080286332
Conscientiousness vs. White Noise -0.138204121
Neuroticism vs. White Noise 0.008313073
Openness vs White Noise 0.150512419
Extroversion vs. Office Noise -0.025252986
Noise Tolerance vs Office Noise 0.103132358
Agreeableness vs Office Noise 0.028077652
83
Conscientiousness vs. Office Noise 0.067395146
Neuroticism vs. Office Noise -0.108630887
Openness vs Office Noise -0.041737765
Average Performance
Extroversion 0.049739119
Noise Tolerance 0.098969352
Agreeableness 0.063082779
Conscientiousness 0.046846631
Neuroticism -0.18311397
Openness 0.020356565
Noise Performance Average (Two tasks with noise
only)
Extroversion 0.080870175
Noise Tolerance 0.090652791
Agreeableness 0.07824096
Conscientiousness -0.046274585
Neuroticism -0.077106758
Openness 0.074581605
Personality/Noise Tolerance Scores vs. Task
Duration by Noise Level
84
Extroversion vs. No Noise -0.0302
Noise Tolerance vs No Noise 0.109929
Agreeableness vs No Noise 0.059981
Conscientiousness vs. No Noise -0.17992
Neuroticism vs. No Noise 0.220944
Openness vs No Noise -0.03097
Extroversion vs. White Noise -0.07958
Noise Tolerance vs White Noise 0.018494
Agreeableness vs White Noise 0.019655
Conscientiousness vs. White Noise 0.174026
Neuroticism vs. White Noise -0.12586
Openness vs White Noise 0.098807
Extroversion vs. Office Noise 0.270089
Noise Tolerance vs Office Noise -0.16615
Agreeableness vs Office Noise 0.029978
Conscientiousness vs. Office Noise -0.13453
Neuroticism vs. Office Noise -0.05926
Openness vs Office Noise 0.121776
Average Overall Duration
Extroversion 0.084098
Noise Tolerance 0.009329
85
Agreeableness 0.079406
Conscientiousness -0.1334
Neuroticism 0.08461
Openness 0.098405
Noise Duration Average (Two tasks with noise
only)
Extroversion 0.143963
Noise Tolerance -0.11084
Agreeableness 0.036727
Conscientiousness 0.026707
Neuroticism -0.13641
Openness 0.163478
Personality/Noise Tolerance
Scores vs. NASA-TLX
MWL by Noise Level
Extroversion vs. No Noise -0.05041
Noise Tolerance vs No Noise -0.00642
Agreeableness vs No Noise -0.01852
Conscientiousness vs. No Noise -0.19956
Neuroticism vs. No Noise 0.156542
Openness vs No Noise 0.004953
Extroversion vs. White Noise -0.27159
Noise Tolerance vs White Noise -0.10899
Agreeableness vs White Noise 0.055841
Conscientiousness vs. White Noise 0.178901
Neuroticism vs. White Noise 0.053029
Openness vs White Noise -0.05244
86
Extroversion vs. Office Noise -0.0707
Noise Tolerance vs Office Noise -0.04356
Agreeableness vs Office Noise -0.00352
Conscientiousness vs. Office Noise 0.027704
Neuroticism vs. Office Noise 0.325507
Openness vs Office Noise 0.109319
MWL Noise Average (Two tasks with
noise only)
Extroversion -0.21891
Noise Tolerance -0.09817
Agreeableness 0.032735
Conscientiousness 0.131378
Neuroticism 0.254648
Openness 0.041328
Personality/Noise Tolerance Scores vs. HRV
Parameters for Data Entry Task
Noise Tolerance vs. LF/HF Data Entry -0.11042
Noise Tolerance vs Mean HRV Data Entry 0.135907
Noise Tolerance vs. Standard Deviation HRV Data Entry 0.100076
Noise Tolerance vs. RMSSD Data Entry 0.041487
Agreeableness vs LF/HF Data Entry -0.24139
Agreeableness vs. Mean HRV Data Entry 0.035521
Agreeableness vs. Standard Deviation HRV Data Entry 0.322372
Agreeableness vs. RMSSD Data Entry 0.370827
Conscientiousness vs LF/HF Data Entry -0.33965
Conscientiousness vs Mean HRV Data Entry -0.28977
Conscientiousness vs Standard Deviation HRV Data Entry 0.120079
Conscientiousness vs RMSSD Data Entry 0.204418
Neuroticism vs. LF/HF Data Entry 0.020921
Neuroticism vs. Mean HRV Data Entry -0.09506
Neuroticism vs. Standard Deviation HRV Data Entry -0.3297
Neuroticism vs RMSSD Data Entry -0.3538
87
Openness vs LF/HF Data Entry -0.05779
Openness vs Mean HRV Data Entry -0.16198
Openness vs Standard Deviation HRV Data Entry -0.2074
Openness vs RMSSD Data Entry -0.18865
Extroversion vs. LF/HF Data Entry -0.02858
Extroversion vs Mean HRV Data Entry -0.1186
Extroversion vs. Standard Deviation HRV Data Entry -0.11725
Extroversion vs. RMSSD Data Entry -0.1
Personality/Noise Tolerance Scores vs. HRV Parameters for
Anomaly Detection Task
Noise Tolerance vs. LF/HF Anomaly Detection -0.29437
Noise Tolerance vs Mean HRV Anomaly Detection 0.173206
Noise Tolerance vs. Standard Deviation HRV Anomaly Detection 0.130326
Noise Tolerance vs. RMSSD Anomaly Detection 0.043917
Agreeableness vs LF/HF Anomaly Detection -0.18109
Agreeableness vs. Mean HRV Anomaly Detection 0.087904
Agreeableness vs. Standard Deviation HRV Anomaly Detection 0.064341
Agreeableness vs. RMSSD Anomaly Detection 0.108255
Conscientiousness vs LF/HF Anomaly Detection -0.11429
Conscientiousness vs Mean HRV Anomaly Detection -0.21318
Conscientiousness vs Standard Deviation HRV Anomaly Detection 0.185376
Conscientiousness vs RMSSD Anomaly Detection 0.080377
Neuroticism vs. LF/HF Anomaly Detection 0.251707
Neuroticism vs. Mean HRV Anomaly Detection -0.05951
Neuroticism vs. Standard Deviation HRV Anomaly Detection -0.23961
Neuroticism vs RMSSD Anomaly Detection -0.22318
Openness vs LF/HF Anomaly Detection 0.042183
Openness vs Mean HRV Anomaly Detection -0.02354
Openness vs Standard Deviation HRV Anomaly Detection -0.15493
Openness vs RMSSD Anomaly Detection -0.15608
Extroversion vs. LF/HF Anomaly Detection -0.03361
88
Extroversion vs Mean HRV Anomaly Detection -0.09883
Extroversion vs. Standard Deviation HRV Anomaly Detection -0.01366
Extroversion vs. RMSSD Anomaly Detection 0.028114
Personality/Noise Tolerance Scores vs. HRV Parameters for
Arithmetic Task
Noise Tolerance vs. LF/HF Arithmetic -0.13366
Noise Tolerance vs Mean Arithmetic HRV 0.115213
Noise Tolerance vs. Standard Deviation HRV Arithmetic 0.051962
Noise Tolerance vs. RMSSD Arithmetic 0.06056
Agreeableness vs LF/HF Arithmetic -0.22496
Agreeableness vs. Mean HRV Arithmetic 0.083366
Agreeableness vs. Standard Deviation HRV Arithmetic 0.136534
Agreeableness vs. RMSSD Arithmetic 0.177938
Conscientiousness vs LF/HF Arithmetic -0.16897
Conscientiousness vs Mean HRV Arithmetic -0.31982
Conscientiousness vs Standard Deviation HRV Arithmetic 0.153044
Conscientiousness vs RMSSD Arithmetic 0.173451
Neuroticism vs. LF/HF Arithmetic 0.015285
Neuroticism vs. Mean HRV Arithmetic -0.11667
Neuroticism vs. Standard Deviation HRV Arithmetic -0.19283
Neuroticism vs RMSSD Arithmetic -0.12771
Openness vs LF/HF Arithmetic -0.05551
Openness vs Mean HRV Arithmetic -0.14122
Openness vs Standard Deviation HRV Arithmetic -0.18512
Openness vs RMSSD Arithmetic -0.17766
Extroversion vs. LF/HF Arithmetic -0.02455
Extroversion vs Mean HRV Arithmetic -0.13699
Extroversion vs. Standard Deviation HRV Arithmetic -0.06578
Extroversion vs. RMSSD Arithmetic -0.12756
Personality/Noise Tolerance Scores vs. HRV Parameters for
No Noise Tasks
Noise Tolerance vs. LF/HF -0.09209
Noise Tolerance vs Mean HRV 0.074328
89
Noise Tolerance vs. Standard Deviation HRV 0.091491
Noise Tolerance vs. RMSSD 0.027043
Agreeableness vs LF/HF -0.34985
Agreeableness vs. Mean HRV 0.008364
Agreeableness vs. Standard Deviation HRV 0.249448
Agreeableness vs. RMSSD 0.281866
Conscientiousness vs LF/HF -0.29021
Conscientiousness vs MEAN -0.35518
Conscientiousness vs HRV Standard Deviation 0.219654
Conscientiousness vs RMSSD 0.228316
Neuroticism vs. LF/HF 0.100183
Neuroticism vs. MEAN -0.08938
Neuroticism vs. HRV Standard Deviation -0.22604
Neuroticism vs RMSSD -0.22593
Openness vs LF/HF 0.070492
Openness vs Mean HRV -0.2037
Openness vs Standard Deviation HRV -0.1922
Openness vs RMSSD -0.18003
Extrovert vs. LF/HF 0.166907
Extrovert vs Mean HRV -0.09624
Extrovert vs. Standard Deviation HRV -0.13203
Extrovert vs. RMSSD -0.15052
Personality/Noise Tolerance Scores vs. HRV Parameters
for White Noise Tasks
Noise Tolerance vs. LF/HF -0.21745
Noise Tolerance vs Mean HRV 0.079274
Noise Tolerance vs. Standard Deviation HRV 0.150001
Noise Tolerance vs. RMSSD 0.115791
Agreeableness vs LF/HF -0.10595
Agreeableness vs. Mean HRV 0.018492
Agreeableness vs. Standard Deviation HRV 0.171613
Agreeableness vs. RMSSD 0.240377
90
Conscientiousness vs LF/HF -0.06272
Conscientiousness vs MEAN -0.26176
Conscientiousness vs HRV Standard Deviation 0.081684
Conscientiousness vs RMSSD 0.168021
Neuroticism vs. LF/HF 0.030842
Neuroticism vs. MEAN -0.04969
Neuroticism vs. HRV Standard Deviation -0.30262
Neuroticism vs RMSSD -0.26189
Openness vs LF/HF -0.10274
Openness vs Mean HRV -0.06779
Openness vs Standard Deviation HRV -0.13624
Openness vs RMSSD -0.13897
Extrovert vs. LF/HF -0.13336
Extrovert vs Mean HRV -0.14759
Extrovert vs. Standard Deviation HRV -0.04475
Extrovert vs. RMSSD -0.2289
Personality/Noise Tolerance Scores vs. HRV Parameters
for Office Noise Tasks
Noise Tolerance vs. LF/HF -0.14037
Noise Tolerance vs Mean HRV 0.156468
Noise Tolerance vs. Standard Deviation HRV 0.075397
Noise Tolerance vs. RMSSD 0.01048
Agreeableness vs LF/HF -0.06766
Agreeableness vs. Mean HRV 0.110398
Agreeableness vs. Standard Deviation HRV 0.13982
Agreeableness vs. RMSSD 0.136357
Conscientiousness vs LF/HF -0.2242
Conscientiousness vs MEAN -0.17513
Conscientiousness vs HRV Standard Deviation 0.16913
Conscientiousness vs RMSSD 0.050387
Neuroticism vs. LF/HF 0.106753
91
Neuroticism vs. MEAN 0.106753
Neuroticism vs. HRV Standard Deviation -0.24382
Neuroticism vs RMSSD -0.21223
Openness vs LF/HF -0.01792
Openness vs Mean HRV -0.05783
Openness vs Standard Deviation HRV -0.23607
Openness vs RMSSD -0.20344
Extrovert vs. LF/HF -0.10355
Extrovert vs Mean HRV -0.14093
Extrovert vs. Standard Deviation HRV -0.12502
Extrovert vs. RMSSD -0.06475
No Noise 0.151738 No Noise -0.08802 No Noise -0.37662
White Noise 0.200556 White Noise -0.00497 White Noise -0.28355
Office Noise 0.158821 Office Noise -0.26256 Office Noise -0.44199
Extroversion vs. Pupil Diameter Difference
(from Control) by Noise Level
Noise Tolerance vs. Pupil Diameter
Difference (from Control) by Noise Level
Agreeableness vs. Pupil Diameter
Difference (from Control) by Noise Level
No Noise 0.134334356 No Noise -0.01491 No Noise 0.24384
White Noise -0.04324164 White Noise -0.12207 White Noise 0.021463
Office Noise -0.00557524 Office Noise 0.022873 Office Noise 0.166714
Openness vs. Pupil Diameter Difference
(from Control) by Noise Level
Conscientiousness vs. Pupil Diameter
Difference (from Control) by Noise Level
Neuroticism vs. Pupil Diameter
Difference (from Control) by Noise Level
No Noise Std.d Dilation 0.072009 No Noise Std.d Dilation -0.10471 No Noise Std.d Dilation 0.117678
White Noise Std.d Dilation 0.101576 White Noise Std.d Dilation -0.18032 White Noise Std.d Dilation -0.00417
Office Noise Std.d Dilation 0.031565 Office Noise Std.d Dilation -0.13536 Office Noise Std.d Dilation -0.03787
Extroversion vs. Standard Deviation (Std.d)
Pupil Diameter by Noise Level
Noise Tolerance vs. Standard
Deviation (Std.d) Pupil Diameter by
Noise Level
Agreeableness vs. Standard Deviation
(Std.d) Pupil Diameter by Noise Level
92
No Noise Std.d Dilation -0.03882 No Noise Std.d Dilation -0.0628 No Noise Std.d Dilation 0.12458
White Noise Std.d Dilation 0.018493 White Noise Std.d Dilation -0.01646 White Noise Std.d Dilation 0.239569
Office Noise Std.d Dilation -0.00855 Office Noise Std.d Dilation -0.08294 Office Noise Std.d Dilation -0.03321
Openness vs. Standard Deviation
(Std.d) Pupil Diameter by Noise Level
Conscientiousness vs. Standard
Deviation (Std.d) Pupil Diameter by
Noise Level
Neuroticism vs. Standard Deviation
(Std.d) Pupil Diameter by Noise Level
Data Entry 0.242012 Data Entry 0.211357 Data Entry -0.14679
Anomaly Detection 0.242012 Anomaly Detection 0.088629 Anomaly Detection -0.1645
Arithmetic 0.306569 Arithmetic 0.163438 Arithmetic -0.09939
Extroversion vs. Pupil Diameter
Difference (from Control) by
Task Type
Noise Tolerance vs. Pupil
Diameter Difference (from
Control) by Task Type
Agreeableness vs. Pupil
Diameter Difference (from
Control) by Task Type
Data Entry 0.033953 Data Entry -0.05129 Data Entry -0.06264
Anomaly Detection -0.09294 Anomaly Detection 0.046656 Anomaly Detection -0.1023
Arithmetic 0.073106 Arithmetic 0.013369 Arithmetic -0.02834
Openness vs. Pupil Diameter
Difference (from Control) by
Task Type
Conscientiousness vs. Pupil
Diameter Difference (from
Control) by Task Type
Neuroticism vs. Pupil Diameter
Difference (from Control) by
Task Type
Data Entry 0.011469 Data Entry 0.116233 Data Entry 0.056845
Anomaly Detection -0.00379 Anomaly Detection 0.249274 Anomaly Detection -0.05663
Arithmetic 0.102957 Arithmetic 0.142628 Arithmetic -0.20096
Extroversion vs. Standard
Deviation (Std.d) Pupil
Diameter by Task Type
Noise Tolerance vs. Standard
Deviation (Std.d) Pupil Diameter by
Task Type
Agreeableness vs. Standard
Deviation (Std.d) Pupil Diameter by
Task Type
Data Entry 0.170295 Data Entry -0.17873 Data Entry 0.209977
Anomaly Detection 0.007025 Anomaly Detection 0.02287 Anomaly Detection 0.100182
Arithmetic 0.04289 Arithmetic 0.06229 Arithmetic 0.141966
Conscientiousness vs.
Standard Deviation (Std.d)
Pupil Diameter by Task Type
Neuroticism vs. Standard
Deviation (Std.d) Pupil
Diameter by Task Type
Openness vs. Standard
Deviation (Std.d) Pupil
Diameter by Task Type
93
8.0 References
Aarde, N., Meiring, D., & Wiernik, B. (2017). The validity of the Big Five personality traits for
job performance: Meta-analyses of South African studies. International Journal of
Selection and Assessment. 25(3). 223-239.
Alansari, B. (2016). The Big Five Inventory (BFI): Reliability and validity of its Arabic
translation in non-clinical sample. European Psychiatry, 33S248. doi:
10.1016/j.eurpsy.2016.01.500
Alessandri, G. and Vecchione, M. (2012). The higher-order factors of the Big Five as predictors
of job performance. Personality and Individual Differences. 53. 779-784.
Arnrich, B., Cinaz, B., Arnrich, B., La Marca, R., & Troester, G. (2011). Monitoring of mental
workload levels during an everyday life office-work scenario. Personal and Ubiquitous
Computing, 17(2), 229-239.
Arterberry, B. J., Martens, M. P., Cadigan, J. M., & Rohrer, D. (2014). Application of
Generalizability Theory to the Big Five Inventory. Personality and Individual
Differences, 6998-103. doi:10.1016/j.paid.2014.05.015
Basil, M. D. (1994). Multiple resource theory I: Application to television viewing.
Communication Research, 21(2), 177.
Becker, A. B., Warm, J. S., Dember, W. N., & Hancock, P. (1995). Effects of jet engine noise
and performance feedback on perceived workload in a monitoring task. International
Journal of Aviation Psychology, 5(1).
94
Belojevic, G., Jakovljevic, B., & Slepcevic, V. (2003). Noise and mental performance:
personality attributes and noise sensitivity. Noise & Health, 6(21), 77-89.
Billman, G. (2013) The LF/HF ratio does not accurately measure cardiac sympatho-vagal
balance. Frontiers in Physiology. 4(26).
Boonghee, Y., Neelankavil, J. P., de Guzman, G. M., & Lim, R. A. (2013). Personality Type
Preferences of Asian Managers: A Cross-Country Analysis Using the MBTI Instrument.
International Journal of Global Management Studies, 5(1), 1-23.
Brotherton, P. (2012). 360 Instruments Are the Most Popular Way to Assess Leadership. T+D,
66(8), 18.
Cail, F., & Aptel, M. (2003). Biomechanical stresses in computer-aided design and in data entry.
International Journal of Occupational Safety and Ergonomics, 9(3), 235-255.
Cardona, G., & Quevedo, N. (2014). Blinking and driving: the influence of saccades and
cognitive workload. Current Eye Research, 39(3), 239-244.
doi:10.3109/02713683.2013.841256
Church, D. G. (2015). Reducing Error Rates in Intelligence, Surveillance, and Reconnaissance
(ISR) Anomaly Detection via Information Presentation Optimization. Wright State
University. Retrieved from
https://etd.ohiolink.edu/pg_10?0::NO:10:P10_ACCESSION_NUM:wright1452858183
Colligan, L., Potts, H. W., Finn, C. T., & Sinkin, R. A. (2015). Cognitive workload changes for
nurses transitioning from a legacy system with paper documentation to a commercial
95
electronic health record. International Journal of Medical Informatics, 84(7), 469-476.
doi:10.1016/j.ijmedinf.2015.03.003
Cooper, C. A., McCord, D. M., & Campbell-Bridges, W. (2017). Personality and the Teaching of
Public Administration: A Case for the Big Five. Journal of Public Affairs Education,
23(2), 677.
Dimitrakopoulos, G. N., Kakkos, I., Dai, Z., Lim, J., deSouza, J. J., Bezerianos, A., & Sun, Y.
(2017). Task-Independent mental workload classification based upon common multiband
EEG cortical connectivity. IEEE transactions on neural systems and rehabilitation
engineering: A publication of the IEEE engineering in medicine and biology society,
25(11), 1940-1949. doi:10.1109/TNSRE.2017.2701002
Dobbs, S., Furnham, A., & McClelland, A. (2011). The effect of background music and noise on
the cognitive test performance of introverts and extraverts. Applied Cognitive
Psychology, 25(2), 307-313.
Dockrell, J., Shield, B. M., & Dockrell, J. E. (2008). The effects of environmental and classroom
noise on the academic attainments of primary school children. Journal of the Acoustical
Society of America, 123(1), 133-144.
Faure, V., Lobjois, R., & Benguigui, N. (2016). The effects of driving environment complexity
and dual tasking on drivers’ mental workload and eye blink behavior. Transportation
Research Part F: Traffic Psychology and Behavior, 40. 78-90.
doi:10.1016/j.trf.2016.04.007
96
Francis, L. J., Lewis, C. A., & Ziebertz, H. (2006). The short-form revised Eysenck personality
questionnaire (EPQR-S): A German edition. Social Behavior & Personality: An
International Journal, 34(2), 197-203.
Francisco Ruiz-Rabelo, J., Navarro-Rodriguez, E., Diaz-Jimenez, N., Cabrera-Bermon, J., Diaz-
Iglesias, C., Gomez-Alvarez, M., & ... Luigi Di-Stasi, L. (2015). Validation of the
NASA-TLX score in ongoing assessment of mental workload during a laparoscopic
learning curve in bariatric surgery. Obesity Surgery, 25(12), 2451-2456.
Furnham, A., & Strbac, L. (2002). Music is as distracting as noise: The differential distraction of
background music and noise on the cognitive test performance of introverts and
extraverts. Ergonomics, 45(3), 203-217. doi:10.1080/00140130210121932
Galy, E., & Mélan, C. (2015). Effects of cognitive appraisal and mental workload factors on
performance in an arithmetic task. Applied Psychophysiology and Biofeedback, 40(4),
313-325. doi:10.1007/s10484-015-9302-0
Gao, Q., Wang, Y., Li, Z., Dong, X., & Song, F. (2013). Mental workload measurement for
emergency operating procedures in digital nuclear power plants. Ergonomics, 56(7),
1070-1085.
Gabbard, R. D. (2017). Identifying the Impact of Noise on Anomaly Detection through
Functional Near-Infrared Spectroscopy (fNIRS) and Eye-tracking. Wright State
University. Retrieved from
https://etd.ohiolink.edu/!etd.send_file?accession=wright1501711461736129&disposition
=inline
97
Gentry, T. A., Wakefield Jr., J. A., & Friedman, A. F. (1985). MMPI Scales for measuring
Eysenck's Personality Factors. Journal of Personality Assessment, 49(2), 146.
Gerhard, M., & Joost de, W. (2015). Workload assessment for mental arithmetic tasks using the
task-evoked pupillary response. Peer Computer Science, 1(16), e16. doi:10.7717/peerj-
cs.16
Gerras, S. J., & Wong, L. (2016). Moving Beyond the MBTI. Military Review, 96(2), 54.
Guo, W., Tian, X., Tan, J., Zhao, L., and Li, L. (2016). Driver's mental workload estimation
based on empirical physiological indicators. 2016 31st Youth Academic Annual
Conference of Chinese Association of Automation (YAC), Chinese Association of
Automation (YAC), Youth Academic Annual Conference of, 344.
doi:10.1109/YAC.2016.7804916
Hart, S. G. (2006). NASA-task load index (NASA-TLX); 20 years later. Paper presented at the
Proceedings of the human factors and ergonomics society annual meeting.
He, X., Wang, L., & Gao, X. & Chen, Y. (2012). IEEE 10th International Conference on
Industrial Informatics, Industrial Informatics (INDIN), 2012 10th IEEE International
Conference on, 502. doi:10.1109/INDIN.2012.6301203
Heine, T., Lenis, G., Reichensperger, P., Beran, T., Doessel, O., & Deml, B. (2017).
Electrocardiographic features for the measurement of drivers' mental workload. Applied
Ergonomics, 6131-43. doi:10.1016/j.apergo.2016.12.015
98
Holmqvist, K., Nyström, M., Andersson, R., Dewhurst, R., Jarodzka, H., & Van de Weijer, J.
(2011). Eye tracking: A comprehensive guide to methods and measures. New York, NY:
Oxford University Press.
Horrey, W.J., & Wickens, C.D. (2003). Multiple resource modeling of task interference in
vehicle control, hazard awareness and in-vehicle task performance. In Proceedings of the
Second International Driving Symposium on Human Factors in Driver Assessment,
Training, and Vehicle Design 2013. 7-12
Hu, J. L., Lu, J., Tan, W. B., & Lomanto, D. (2016). Training improves laparoscopic tasks
performance and decreases operator workload. Surgical Endoscopy, 30(5), 1742-1746.
doi:10.1007/s00464-015-4410-8
Hygge, S., & Knez, I. (2001). Effects of noise, heat, and indoor lighting on cognitive
performance and self-reported affect. Journal of Environmental Psychology, 21. 291-299.
doi:10.1006/jevp.2001.0222
Jafari, M. J., & Kazempour, M. (2013). Mental processing of human subjects with different
individual characters exposed to low frequency noise. International Journal of
Occupational Hygiene, 5(3), 64-70.
Jahncke, H., Björkeholm, P., Marsh, J. E., Odelius, J., & Sörqvist, P. (2016). Office noise: Can
headphones and masking sound attenuate distraction by background speech? Work, 55(3),
505-513. doi:10.3233/WOR-162421
Jerison, H. J. (1959). Effects of noise on human performance. Journal of Applied Psychology,
43(2), 96-101. doi:10.1037/h0042914
99
John, O. P., Naumann, L. P., & Soto, C. J. (2008). Paradigm shift to the integrative Big-Five trait
taxonomy: History, measurement, and conceptual issues. In O. P. John, R. W. Robins, &
L. A. Pervin (Eds.), Handbook of personality: Theory and research (pp. 114-158). New
York, NY: Guilford Press.
Judge, T., Rodell, J., Klinger, R., and Simon, L. (2013) Hierarchical representations of the Five-
Factor Model of Personality in predicting job performance: Integrating three organizing
frameworks with two theoretical perspectives. Applied Psychology. 98(6). 875-925.
Khan, Z. A., & Rizvi, S. H. (2010). A study on the effects of human age, type of computer and
noise on operators' performance of a data entry task. International Journal of
Occupational Safety and Ergonomics, 16(4), 455-463.
Kotani, K., Takamasu, K., & Tachibana, M. (2007). Respiratory-phase domain analysis of heart
rate variability can accurately estimate cardiac vagal activity during a mental arithmetic
task. Methods of Information In Medicine, 46(3), 376-385.
Kou, S., Furnham, A., McClelland, A., & Furnham, A. (2018). The effect of background music
and noise on the cognitive test performance of Chinese introverts and extraverts.
Psychology of Music, 46(1), 125-135.
Lahtela, K., Niemi, P., Kuusela, V., & Hypén, K. (1986). Noise and visual choice-reaction time:
A large-scale population survey. Scandinavian Journal of Psychology, 27(1), 52-57.
doi:10.1111/j.1467-9450.1986.tb01186.x
100
Lercher, P., Evans, G. W., & Meis, M. (2003). Ambient noise and cognitive processes among
primary schoolchildren. Environment and Behavior, 35(6), 725-735.
doi:10.1177/0013916503256260
Levy-Leboyer, C. (1989). Noise effects on two industrial tasks. Work & Stress, 3(4), 315-322.
doi:10.1080/02678378908256949
Liang, S. M., Rau, C., Tsai, P., & Chen, W. (2014). Validation of a task demand measure for
predicting mental workloads of physical therapists. International Journal of Industrial
Ergonomics, 44. 747-752. doi:10.1016/j.ergon.2014.08.002
Ljungberg, J. K., & Neely, G. (2007). Stress, subjective experience and cognitive performance
during exposure to noise and vibration. Journal of Environmental Psychology, 2744-54.
doi:10.1016/j.jenvp.2006.12.003
Lovik, A., Verbeke, G., & Molenberghs, G. (2017). Evaluation of a very short test to measure
the big five personality factors on a Flemish sample. Journal of Psychological &
Educational Research, 25(2), 7-17.
Marquart, G., Cabrall, C., & de Winter, J. (2015). Review of eye-related measures of drivers’
mental workload. Procedia Manufacturing, 3(6th International Conference on Applied
Human Factors and Ergonomics (AHFE 2015) and the Affiliated Conferences, AHFE
2015), 2854-2861. doi:10.1016/j.promfg.2015.07.783
Mansikka, H., Virtanen, K., Harris, D., and Simola, P. (2016). Fighter pilots’ heart rate, heart
rate variation and performance during an instrument flight rules proficiency test. Applied
Ergonomics. 56. 213-219.
101
Myers, S. (2016). Myers-Briggs typology and Jungian individuation. Journal of Analytical
Psychology, 61(3), 289-308. doi:10.1111/1468-5922.12233.
Nassiri, P., Monazam, M., Fouladi Dehaghi, B., Abadi, L., Zakerian, SA., & Azam, K. (2013).
The effect of noise on human performance: A clinical trial. The International Journal of
Occupational and Environmental Medicine, 4(2), Pp 87-95 (2013), (2), 87.
Nickels, M. (2014). IMPROVING MOTION IMAGERY ANALYSIS: INVESTIGATING
DETECTION FAILURES, REMEMBERING TO PERFORM DEFERRED INTENTIONS.
(Electronic Thesis or Dissertation). Retrieved from https://etd.ohiolink.edu/
Othman, N. and Romli, F. (2016). Mental workload evaluation of pilots using pupil dilation.
International Review of Aerospace 9(3).
Peng, X., He, Q., Ji, T., Wang, Z., & Yang, L. (2006). [Mental workload for mental arithmetic on
visual display terminal]. Zhonghua Lao Dong Wei Sheng Zhi Ye Bing Za Zhi = Zhonghua
Laodong Weisheng Zhiyebing Zazhi = Chinese Journal of Industrial Hygiene and
Occupational Diseases, 24(12), 726-729.
Pittenger, D. J. (1993). The Utility of the Myers-Briggs Type Indicator. Review of Educational
Research, (4), 467.
Pittenger, D. J. (2005). Cautionary Comments Regarding the Myers-Briggs Type Indicator.
Consulting Psychology Journal: Practice & Research, 57(3), 210-221. doi:10.1037/1065-
9293.57.3.210
102
Piasecki, A. M. (2016). Improving Anomaly Detection through Identification of Physiological
Signatures of Unconscious Awareness. Wright State University. Retrieved from
https://etd.ohiolink.edu/pg_10?0::NO:10:P10_ACCESSION_NUM:wright1461764365
Prabhu, A., Smith, W., Yurko, Y., Acker, C., & Stefanidis, D. (2010). Simulation-Based Surgical
Education: Increased stress levels may explain the incomplete transfer of simulator-
acquired skill to the operating room. Surgery, 147640-645.
doi:10.1016/j.surg.2010.01.007
Pujol, S., Levain, J., Houot, H., Petit, R., Berthillier, M., Defrance, J., & ... Mauny, F. (2014).
Association between Ambient Noise Exposure and School Performance of Children
Living in An Urban Area: A Cross-Sectional Population-Based Study. Journal of Urban
Health, 91(2), 256-271. doi:10.1007/s11524-013-9843-6.
Rabinowitz P.M. 2000. Noise-induced hearing loss. American Family Physician, 61(9),
2749–2759.
Rammstedt, B. & John, O. (2007). Measuring personality in one minute or less: A 10-item short
version of the Big Five Inventory in English and German. Journal of Research in
Personality. 41(1). 203-212.
Reyes Zamorano, E., Álvarez Carrillo, C., Peredo Silva, A., Sandoval, A. M., & Rebolledo
Pastrana, I. M. (2014). Psychometric properties of the big five inventory in a Mexican
sample. Salud Mental, 37(6), 491-497.
103
Rodrigues, N., & Rebelo, T. (2013). Incremental validity of proactive personality over the Big
Five for predicting job performance of software engineers in an innovative context.
Revista De Psicologia Del Trabajo Y De Las Organizaciones, 29(1), 21-27.
doi:10.5093/tr2013a4
Rubio, S., Díaz, E., Martín, J., & Puente, J. M. (2004). Evaluation of subjective mental
workload: A comparison of SWAT, NASA-TLX, and workload profile methods. Applied
Psychology: An International Review, 53(1), 61-86. doi:10.1111/j.1464-
0597.2004.00161.x
Rummel, R. J. (1976). Understanding Correlation. Retrieved from
http://www.hawaii.edu/powerkills/UC.HTM
Sönmez, B., Oğuz, Z., Kutlu, L., & Yıldırım, A. (2017). Determination of nurses' mental
workloads using subjective methods. Journal of Clinical Nursing, 26(3-4), 514-523.
doi:10.1111/jocn.13476
Spector, R. (1990). Chapter 58: The Pupils. In H.K. Walker, W.D. Hall, & J.W. Hurst (Ed. 3),
Clinical Methods: The History, Physical, and Laboratory Examinations. Boston:
Butterworths.
Sugimoto, I., Kitamura, K., Murai, K., Wang, Y., and Wang, J. (2016) Study on relation between
operator and trainee's mental workload for ship maneuvering simulator exercise using
heart rate variability. 2016 IEEE International Conference on Systems, Man, and
104
Cybernetics (SMC) Systems, Man, and Cybernetics (SMC), 2016 IEEE International
Conference, Budapest, Hungary, Oct. 2016.
Szalma, J. L., & Hancock, P. A. (2011). Noise effects on human performance: A meta-analytic
synthesis. Psychological Bulletin, 137(4), 682-707. doi:10.1037/a0023987
Tafalla, R. J., & Evans, G. W. (1997). Noise, physiology, and human performance: The potential
role of effort. Journal of Occupational Health Psychology, 2(2), 148-155.
doi:10.1037/1076-8998.2.2.148
Tokuda, S., Obinata, G., Palmer, E., & Chaparro, A. (2011). Estimation of mental workload
using saccadic eye movements in a free-viewing task. Conference Proceedings: Annual
International Conference Of The IEEE Engineering In Medicine And Biology Society.
IEEE Engineering In Medicine And Biology Society. Annual Conference, 20114523-
4529. doi:10.1109/IEMBS.2011.6091121
Van Orden, K. F., Limbert, W., Makeig, S., & Jung, T. (2001). Eye activity correlates of
workload during a visuospatial memory task. Human Factors, 43(1), 111-121.
doi:10.1518/001872001775992570
Vollmer, M. (2015). A robust, simple and reliable measure of heart rate variability using relative
RR intervals. 2015 Computing In Cardiology Conference (Cinc), 609.
doi:10.1109/CIC.2015.7410984
Weinstein, N. D. (1974). Effect of noise on intellectual performance. Journal of Applied
Psychology, 59(5), 548-554.
105
Wickens, C., & Wickens, C. D. (2008). Multiple resources and mental workload. Human
Factors, 50(3), 449-455.
Wickens, C. D. (2002). Multiple resources and performance prediction. Theoretical Issues in
Ergonomics Science, 3(2), 159-177. doi:10.1080/14639220210123806
Wickens, C. Lee, J.D., Liu, Y., and Becker, S.E.G. (2004). An Introduction to Human Factors
Engineering, 2nd Edition, New Jersey, Prentice Hall.
Top Related